One-semester projects

This is a list of topics for one-semester project for visiting students. Those who are interested should contact the supervisor by email well in advance to discuss detailed specification. The student will be registered only after approval of the supervisor. Number of credits will be allocated individually according to each specific project. Some of the projects are also available for interns during Summer.

Department:
Supervisor:
Branch of study:
Intended for:
Topic
Supervisor
Duration
Department
Date
Deep spatio-temporal models for detecting changes of land cover use.
doc. Boris Flach, Dr. rer. nat. habil.
1 semester / Master
Department of Cybernetics
11. 7. 2018
Two possible diploma topics in the context of a project funded by the European Space Agency in collaboration with Mapradix s.r.o. The project aims at spatio-temporal probabilistic neural models for classifying and detecting changes of land cover use in satellite images. The models and approaches will use the following types of initial data: medium spatial resolution and high temporal resolution satellite data as well as temporally sparse but very high spatial resolution satellite & airborne data. All data are geo-referenced and registered. Possible topics are: 1. Novel deep recurrent models for spatially coupled time processes: Develop novel recurrent deep networks for modelling spatially coupled time processes. Apply them on the medium spatial resolution and high temporal resolution data for classification of land cover use and detection of changes. 2. Combining deep networks and graphical models for super-resolution transfer: Given spatio-temporal classification of land cover use in medium spatial resolution and a single high resolution satellite image, transfer the segmentation from the medium to the high resolution. This topic aims at combining deep CNNs with graphical models (Markov Random Fields) in order to achieve high quality super resolution transfer.
Electronic Circuit Analysis in Maple program
doc. Dr. Ing. Jiří Hospodka
1 semester / All students
Department of Circuit Theory
25. 5. 2018
The project is focused on testing of simulation software for electronic circuit analysis. The simulation software consists of a special package named PraCAn for program Maple. The goal of the project is to develop test scripts for the entire package and also examples on which the rightness and accuracy of the calculations will be tested. As a comparison simulator we use WinSpice program, but another type of simulator (PSpice, MicroCap, ...) can be used.

Requirements

Knowledge of basic methods for electric and electronic circuit analysis, knowledge of a Spice simulator and basic programming skills. Knowledge of the Maple program and its programming language is an advantage. Recommended subjects to be studied at the DEF: Analog Circuits (available for fall semester) Laboratory of Electronic Systems (spring semester) Integrated System Structures (fall semester) Implementation of Analog Systems (fall semester)
Statistical model of line of sight shadowing of GNSS satellites
doc. Dr. Ing. Pavel Kovář
1 semester / Doctoral
Department of Radioelectronics
11. 4. 2017
The project aims is to develop a statistical model of a shadowing of a line of sight of the GNSS satellites for various realistic environments. The theoretical part will cover selection of an appropriate stochastic process for modeling of the satellite shadowing and its implementation to the MATLAB or MATLAB Simulink. The experimental part will cover extensive data collection campaign that will serve for model parameters adjustment and model verification. The developed model will serve for generation of the simulation scenario for the GNSS simulator.

Requirements

Strong basics in mathematics, physic, signal processing, radio engineering, satellite navigation and knowledge of MATLAB and Simulink.
GridCut integration
doc. Ing. Daniel Sýkora, Ph.D.
1 semester / Master
Department of Computer Graphics and Interaction
11. 4. 2017
Study an existing software library GridCut for efficient computation of maximum flow on graphs with grid-like topology. Integrate GridCut into the following software packages: Matlab, OpenCV, and GCO. Compare its performance against built-in solvers on the following practical applications: image segmentation, digital photomontage, stereo reconstruction, and multi-label optimization.

Requirements

Basic skills in Matlab, OpenCV, and GCO.
Calibration Platform for Navigation Systems
doc. Ing. Jan Roháč, Ph.D.
1 semester / All students
Department of Measurement
25. 5. 2018
The project is dedicated to a design and development ofa calibration platform using Pan&Tilt device PTU D46-70 (FLIR manufacturer) for calibration of navigation systems in terms of inertial sensors, such as 3 axial accelerometersand 3 axial gyroscopes. The calibration should be at the end fully autonomous. Even if the assignment includes a design and development of both HW and SW, the emphasis is put on SW. The SW should control required maneuvers with the platform, pool the data and estimate an error model of the navigation system being calibrated. Optimally all should be done in MATLAB environment. The project develops deep knowledge of navigation systems applicable in aerial and/or terrestrial vehicles including navigation systems’ calibration, sensors and their stochastic/deterministic errors compensation/modeling, Allan variance analyses.

Requirements

Matlab.Knowledge of calibration procedures, sensor error modeling, navigation principles will be useful, but it is not required. All can be a part of consultations.
Indoor navigation
doc. Ing. Jan Roháč, Ph.D.
1 semester / Master
Department of Measurement
11. 7. 2018
The project is focused to design and develop SW solution for an indoor navigation system primary consisting of IMU (inertial measurement unit – accelerometers and gyroscopes) and supplemented by a multi-camera system. The navigation solution will use extended Kalman filtering approach to fuse data from available information sources in order to provide stable estimation of position, velocity, and attitude. The project will cover computer vision techniques implemented to obtain relative positioning information which will further be fused with inertial data. The solution should be confirmed experimentally.

Requirements

Matlab. Knowledge of Kalman filters will be useful, but it is not required. All can be a part of consultations.
Navigation of Unmanned Vehicles with an Indirect Extended Kalman Filter
doc. Ing. Jan Roháč, Ph.D.
1 semester / All students
Department of Measurement
12. 4. 2017
Design and develop a model applicable in Extended Kalman Filter (EKF) to fuse inertial data (acceleration, angular rates) with GNSS based positioning. The model is supposed to be in an indirect form and should enable the estimation of errors in the position, linear velocity, and attitude. The estimation should be robust to GNSS signal outages and provide navigation as standalone at least within a minute without GNSS aiding. A model conditioning should be analyzed and set with respect to the best performance in a standalone mode. Verify the model on real flight data and compare the results to already an existing direct form of the EKF. Study the model behavior in terms of biases and adaptive conditioning. The project develops deep knowledge of navigation systems applicable in aerial and/or terrestrial vehicles including navigation systems’ calibration, sensors and their stochastic/deterministic errors compensation/modeling, Allan variance analyses, data fusion and preprocessing.

Requirements

Matlab. Knowledge of Kalman filters will be useful, but it is not required. All can be a part of consultations.
UWB localization system
doc. Ing. Jan Roháč, Ph.D.
1 semester / Master
Department of Measurement
11. 7. 2018
The project aims a design and development of UWB indoor localization system enabling to track and localize targets indoor on two-way-ranging (TWR) principle. The project covers HW design of the system, its programming and verification. The solution will substitute then GNSS in indoor environment which will increase overall indoor navigation performance and accuracy of navigated platforms/vehicles. This technology is promising since it provides better accuracy than Wi-Fi indoor navigation localization systems.

Requirements

HW design, C language programming
Virtual sensors in navigation systems
doc. Ing. Jan Roháč, Ph.D.
1 semester / Master
Department of Measurement
11. 7. 2018
The project is focused to application of virtual sensors in navigation solutions. It covers a design and development of algorithms enabling utilization of perfect measurements in terms of virtual sensors. The design should be verified on experiments already taken on a fast boat operated in Norwegian fjords and on a car in urban areas. The robustness, stability, and accuracy are a main target.

Requirements

Matlab. Knowledge of Kalman filters will be useful, but it is not required. All can be a part of consultations.
New Reliable and Efficient Algorithms for Electronic Circuit Optimization
doc. Ing. Josef Dobeš, CSc.
1 semester / Doctoral
Department of Radioelectronics
25. 5. 2018
The aim of the project is to compare (and improve) contemporary algorithms for electronic circuit optimization. Generally, optimization of electronic circuits in DC, frequency, and time domains is a difficult task due to big complexity of the problem: large system matrices, a lot of variables (sometimes hundreds) for the optimization, a number of physical constraints, and large differences between circuit variables/parameters which lead to highly unbalanced matrices. Therefore, the task consists in selecting, deriving, and improving algorithms, programming them to our circuit analyzer, and compare them using our library of typical circuits.

Requirements

Good knowledge of methods for computer-aided design of electronic circuits, basic skills on mathematical optimization, and to be experienced in C/C++/Fortran programming.
Switched-mode high power current source
doc. Ing. Julius Foit, CSc.
1 semester / Doctoral
Department of Microelectronics
12. 4. 2017
The project is focused at the design of a high-efficiency overvoltage-protected high-power current
source: nominal output current should be 10 A 5%; operating temperature range -10 to +55 0C;
relative humidity 0 to 100% (splashing fresh-water resistant); switched-mode operation should be
used, targeted at highest possible efficiency at output voltages from 5.5 to 14.50 V; the output current
should not exceed 25 mA at output voltages over 14.60 V; open-circuit output voltage (i.e. at zero
output current) should not exceed 14.8 V; output current ripple less than 10% peak to peak at all
output voltages up to 14.5 V; external power source: standard 230 V 10%, 50-60 Hz mains; indicator
lights should be provided, signaling: “power on”, “excessive output voltage”, “operation failure”

Requirements

Requirements: Knowledge of basic operational principles of switched-mode power sources, of the properties of small-signal and high-power semiconductor devices including IC´s, principles of design of pulse power transformers, principles of application of temperature-independent reference voltage sources, principles of voltage-to-current conversion. Recommended subjects to be studied at the DEF: Design of Power Supplies for Electronics AE0M34NNZ (available for fall semester) Electron devices AE2B34ELP (available for spring semester) Smart Electronics AE2B34IAA (available for fall semester) References (recommended literature related to the project): [1] P. Horowitz, W. Hill: The Art of Electronics; Cambridge University Press [2] J. Foit: Electronics Fundamentals; Česká technika – nakladatelství ČVUT (CTU Press) further literature according to consultation with the supervisor.
Optimization of Lead-Free Soldering Process
doc. Ing. Karel Dušek, Ph.D.
1 semester / Doctoral
Department of Electrotechnology
11. 4. 2017
Lead-free soldering, which substituted tin-lead soldering in accordance with the RoHS EU Directive, is joined with many problems. Properties of joints differ for different types of surface finishes of joined parts, are strongly dependent on the type and parameters of a soldering process and on properties of lead-free solder alloys. The goal of the work is to optimize a soldering process (vapor soldering, soldering in a tunnel oven) parameters for more types of lead-free solders and three types of surface finishes to form a lead-free joint with electrical and mechanical properties comparable with properties of joints formed of tin-lead solders.
The task is directed to:
1. Analyzis of parameters of solders and soldering process, which dominantly influence parameters of soldered joints formed of lead-free solders.
2. Design of layout of test boards with three types of pads surface finishes (Cu, ENIC, HAL) for measurement of electrical parameters of soldered joints.
3. Forming of test samples of solder joints by SMT assembly of jumpers on the test boards. The joints will be formed of selected types of lead-free solders on three types of surface finishes (Cu, ENIC and HAL). Different parameters and types of asoldering process (vapor soldering, soldering in a tunneloven) will be used.
4. Measuring of electrical and mechanical parameters of solder joints.
5. Discussion of results.

Requirements

Packaging in electronics, electrically conductive jointing in electrical engineering.
ADCS system for a CubeSat
doc. Ing. Martin Hromčík, Ph.D.
1 semester
Department of Control Engineering
12. 4. 2017
The project is devoted to development and testing of an ADCS for a developed CTU CubeSat (CzechTechSat). DCE is responsible for the control laws development and verification, and for related HW and SW problems, and also for sensors fusion and processing tasks. The student would join an existing small team and depending on his or her preferences, and also on the status of the project, shall work on simulation and verification problems, new algorithms for data processing, target platform implementation of selected procedures, design of new control algorithms, etc.

Requirements

Bc. or MSc. student with background in systems and controls.
Subsystems and solutions for small UAV: development and verification
doc. Ing. Martin Hromčík, Ph.D.
1 semester / Bachelor
Department of Control Engineering
25. 5. 2018
The project is focused on further development of a students' UAV platform developed at DCE. Depending on the student's knowledge, skills and interests, a task related to mathematical model construction and refinement can be assigned, as well as advanced control laws design problems, or, for more HW and SW oriented applicants, related programming or construction issues can be addressed (user interface development, ground station components, new sensors and boards incorporation).

Requirements

student with background in systems and controls Suitable supporting courses at CTU: Control Systems for Aircraft and Spacecraft (CSAS), Automatic Control (ARI)
Modal Analysis of Electrically Small Radiators
doc. Ing. Pavel Hazdra, Ph.D.
1 semester
Department of Electromagnetic Field
12. 4. 2017
Suitable for: Ph.D. student, (advanced MS. Student)

The project is aimed on characteristic mode analysis of small antennas/radiators. These very
illustrative elementary solutions are obtained by decomposition of the Method of Moments
impedance matrix. DEF has number of in-house MATLAB codes including mesh generator,
MoM code, characteristic mode analyzator etc. The challenging study will involve deep analysis
of topological properties of small antennas related to radiation efficiency, radiation Q-factor, bandwidth and gain.

Requirements

Strong basics in the area of electromagnetics and antenna theory, knowledge of MATLAB and some EM simulation software (CST/HFSS/FEKO...)
Large Vocabulary Continous Speech Recognition for Multiple Languages
doc. Ing. Petr Pollák, CSc.
1 semester / All students
Department of Circuit Theory
25. 5. 2018
Within this project student should update available large vocabulary continous speech recognizer (LVCSR). The possible update could cover an improvement of acoustic modelling, language modelling, or inclusion of new language depending on available data. Updated recognizer should be tested using speech data from available speech databases. Recognizer should be realized using publicly available KALDI Toolkit.

Requirements

Student should be familier with the basis of digital signal processing, some experiences in speech processing are invited. Recommended subjects for study at CTU FEE are AE2M99CZS Digital Signal Processing (taught in winter semester) and AE2M31ZRE Speech processing (taught in summer semester). The passing of these subjects is not mandatory, it is a recommended option, especially for less experiences students. Basic experience in programming and with operation system Linux is strognly invited.
Small Vocabulary Speech Recognition for Selected Languages
doc. Ing. Petr Pollák, CSc.
1 semester / All students
Department of Circuit Theory
25. 5. 2018
Within this project student should realize small vocabulary speech recognizer for a chosen language, e.g. digit recognizer (for voiced controlled dialing), command recognizer (voiced control of a consumer device), name recognizer (voiced controlled information system). Recognizer should be realized using publicly available KALDI Toolkit. Currently it could be done for English, Spanish, German, French, Portuguese, Slovak, Polish, Hungarian, and Russian. Accuracy of recognition should be tested on defined evaluation set which preparation could also be a part of the project.

Requirements

Student should be familier with the basis of digital signal processing, some experiences in speech processing are invited. Recommended subjects for study at CTU FEE are AE2M99CZS Digital Signal Processing (taught in winter semester) and AE2M31ZRE Speech processing (taught in summer semester). The passing of these subjects is not mandatory, it is a recommended option, especially for less experiences students. Basic experience in programming and with operation system Linux is strognly invited.
Combinatorial algorithms acceleration on GPUs
doc. Ing. Přemysl Šůcha, Ph.D.
1 semester / Doctoral
Department of Control Engineering
25. 5. 2018
The aim of the project is development of parallel algorithms for a combinatorial problem (e.g. Resource Constraint Project Scheduling Problem) using the GPU (Graphics Processing Unit). The algorithm design will be done using CUDA framework. The particular aim would be to find the optimal solution of the problem (e.g. by branch and bound method) using several GPU carts within a single host. Therefore significant part of the work is design of a suitable GPU – GPU communication allowing efficient solution space searching

Requirements

knowledge of C language and algorithm development.
Task mapping and scheduling using column generation
doc. Ing. Přemysl Šůcha, Ph.D.
1 semester / Doctoral
Department of Control Engineering
25. 5. 2018
This project deals with off-line mapping and scheduling of tasks on a network of computational nodes. Specifically, the issue is the optimal mapping of a task-graph representing an algorithm on a network of distributed nodes. The goal of this project is to propose a mathematical model of the problem and solve it using column generation, i.e. a method able to solve large-scale integer linear programming models.

Requirements

student with knowledge of mixed integer linear programming and C++ language.
Design, fabrication and investigation properties of polymer optical multimode planar waveguides
doc. Ing. Václav Prajzler, Ph.D.
1 semester / Doctoral
Department of Microelectronics
12. 4. 2017
The project is aimed to the design, fabrication and investigation properties of polymer
multimode planar optical waveguides. The polymer waveguides will be used for design
fabrication and measurement of polymer 1 x 2 multimode optical power splitter with
input/output standard plastic optical fiber. This component has huge potential in short-distance
communication for applications such as in automobile networks, private office and home
network.

Requirements

Requirements: Experience with Beam Propagation method and Ray tracing advantage but no necessity. Recommended subjects to be studied at the CTU: Optical Systems and Networks Optical Radiation Detection and Detectors Technology of Optical Devices References (the project is closely related to work presented in the following papers): [1] KLOTZBUECHER, T., BRAUNE, T., DADIC, D., SPRZAGALA, M., KOCH, A. Fabrication of optical 1x2 POF splitters using the Laser-LIGA technique, In Proceedings Laser Micromachining for Optoelectronic Device Fabrication, 2003, vol. 4941, p. 121-132. [2] MIZUNO, H., SUGIHARA, O., JORDAN, S., OKAMOTO, N., OHAMA, M., KAINO, T. Replicated polymeric optical waveguide devices with large core connectable to plastic optical fiber using thermo-plastic and thermocurable resins. Journal of Lightwave Technology, 2006, vol. 24, no. 2, p. 919-926. [3] PARK, H.J., LIM, K.S., KANG, H.S. Low-cost 1×2 plastic optical beam splitter using a V-type angle polymer waveguide for the automotive network. Optical Engineering, 2011, vol. 50, no. 7, p. 075002-075004. [4] KWANG, T.K., MIN, K.K. Kuk Low-loss 1 x 2 plastic optical fiber coupler incorporating a tapered polymeric waveguide and plastic optical fiber transition Regions. Japanese Journal of Applied Physics, 2012, vol. 51, no. 9, p. 8504-8506. [5] PRAJZLER, V., PHAM, N.K., ŠPIRKOVÁ, J. Design, fabrication and properties of the multimode polymer planar 1 x 2 Y optical splitter. Radioengineering, 2012, vol. 21, p. 1202-1207.
Texture Parameterization for Large Polygonal Model
doc. Ing. Vlastimil Havran, Ph.D.
1 semester / Master
Department of Computer Graphics and Interaction
11. 4. 2017
Study the methods and applications used for mapping of 2D coordinates to 3D polygonal models. Select the suitable and efficient approach for texture mapping of large model. Provide the texture mapping for a particular model of MPI Informatics Building Model and render example images in PBRT to verify the correctness of the resulting parameterization.

Requirements

Basic skills in computer graphics andtexture mapping.
Allocation of computing and communication resources for Mobile Edge Computing
doc. Ing. Zdeněk Bečvář, Ph.D.
1 semester / Master or doctoral
Department of Telecommunications Engineering
11. 7. 2018
Develop an algorithm allocating resources to multiple users considering their computing requirements and their movement. Test performance of the proposed solution by simulations or in an emulated mobile network.

Requirements

Basics of channel modelling, basic principles of communication, MATLAB and/or C
Drones in mobile networks
doc. Ing. Zdeněk Bečvář, Ph.D.
1 semester / Master or doctoral
Department of Telecommunications Engineering
11. 7. 2018
The topic is focused on integration of drones into future mobile networks in order to improve quality and efficiency of the communication. There are several possible objectives and sub-topics including, for example: propose an algorithm, which selects position of the flying base station (base station carried by a drone) in real-time according to the user's requirements on communication capacity and according to energy consumption; or implement algorithm for positioning of the flying base station into our developed test-bed and perform tests validating implementation,; or design an algorithm for deployment of the flying base stations to support communication among vehicles; etc.

Requirements

Basics of channel modelling, basic principles of communication, MATLAB and/or C
Evaluation of methods for measurement of a position on a planar surface
doc. Ing. Zdeněk Hurák, Ph.D.
1 semester
Department of Control Engineering
12. 4. 2017
The goal of this student project is to investigate and evaluate several methods for
measurement of a position of a single or multiple objects in a small planar arena (say, up to
1m2). The task is motivated by the currently ongoing research in planar feedback
manipulation such as positioning an iron ball on top of a rectangular array of electromagnets.

The particular list of methods that should be studied, evaluated and compared is:
• processing the images taken by a camera observing the global scene (from above)
• resistive foil
• LED matrix
• set of IR sensors
• capacitive sensors
• magnetic sensors .
As a continuation of this task, some development of algorithms for the chosen hardware
platform is expected.

Requirements

Student profile: senior undergraduate or graduate student in electrical engineering. The task needs a creative student with some modest hobby-level skills in electronics. These could be developed while working on the project, of course. The willingness to learn these hardware oriented skills is crucial.
Simulation of nanostructures
doc. RNDr. Jan Voves, CSc.
semester / All students
Department of Microelectronics
25. 5. 2018
The project is focused on the quantum based simulation of electronic and phonon structure and transport properties of nanometric objects with application to new electron devices. The commertial atomistic models will be used. The results will be compared with the experimental data from transport measurement and with the measured Raman spectra. Simulation results will be used for the structure design optimisation.

Requirements

Requirements: Strong basics in the area semiconductor devices and physics, knowledge of TCAD simulation tools like ATLAS or Synopsys is an advantage. Recommended subjects to be studied: TCAD Aplication XP34AT, Electron Transport in Semiconductors XP34ETS, Solid State Physics BE2M34SST. References (the project is closely related to work presented in the following papers): [1] Bayani, A. H.; Dideban, D.; Voves, J.: Investigation of sub-10nm cylindrical surrounding gate germanium nanowire field effect transistor with different cross-section areas, Superlattices and Microstructures, vol. 105, p. 110-116, 2017, [2] Davydova, M., Laposa, A., Smarhak, J., Kromka, A., Neykova, N., Nahlik, J., Kroutil, J., Drahokoupil, J., Voves, J.: Gas sensing behaviour of ZnO/diamond nanostructures, Beilstein J. of Nanotechnology, vol. 9, pp. 22-29, 2018, [3] Bayani, A. H., Voves, J., Dideban, D.: Effective mass approximation versus full atomistic model to calculate the output characteristics of a gate-all-around germanium nanowire field effect transistor (GAA-GeNW-FET), Superlattices and Microstructures, vol. 133, pp. 769-776, 2018.
Multiagent planning
Ing. Antonín Komenda, Ph.D.
1 semester
Department of Computer Science and Engineering
12. 4. 2017
The principles of local search in planning graphs was used in top performing planner LPG in International Planning Competition 2004. Nowaday most efficient single-agent planners use the planning graph stucture for relaxation heuristic estimation. Particular ways of usage of planning graph for local search in multi-agent setting is an interesting unanswered research question currently.
In the assignment,
1) study the principle of local search in planning graph,
2) extend the work on LGP towards distributed search,
3) reuse parts of LPG and complete an implementation of a multi-agent LPG,
4) experimentally verify and validate efficiency of the planner against state-of-the-art solutions (e.g., multi-agent A* and multi-agent GraphPlan).
Database for IVE tool
Ing. Ivo Malý, Ph.D.
semester / All students
Department of Computer Graphics and Interaction
25. 5. 2018
IVE tool is integrated visualization environment for analysis of usability test data. It allows analysis of data sets from different types of data sources using various visualization plugins. Data sets are currently stored in local object database. Analyze formats of data sets and types of databases and suggest new type of database for client-server solution and implement the solution. IVE tool is based on NetBeans Platform and implemented in Java.

Requirements

Java, databases.
Multimodal tablet application for exploration of small indoor spaces for visually impaired children
Ing. Jan Balata
1 semester
Department of Computer Graphics and Interaction
11. 4. 2017
Analyze research dealing with orientation and spatial memory of blind children. Design and develop a multimodal application for touch screen tablets which helps blind children to explore and learn small indoor spaces (e.g. new nursery school). Use voice, sound, touch, gestures, etc. for interaction.

Requirements

Basic skills in user interface design.
Determination of hemodynamic parameters and primary screening of atherosclerosis
Ing. Jan Havlík, Ph.D.
1 semester
Department of Circuit Theory
12. 4. 2017
Suitable for: Ph.D. student (or advanced MSc. student)
The project deals with the design and implementation of signal processing methods for
measuring of blood pressure and hemodynamic parameters.
Consequently, the target of the project is a design of methods for
primary screening of atherosclerosis based on hemodynamic parameters and a shape of the pulse wave.
Anexperimental workplace has been designed and developedfor solving theproject.
New signal database hasbeen collecting using this workplace.

Requirements

Strong basics in the area of signal processing and/or circuit theory and hardware design, knowledge of MATLAB
Inductive wireless power transfer
Ing. Jan Kraček, Ph.D.
1 semester / All students
Department of Electromagnetic Field
11. 7. 2018
The project is aimed at the investigation of inductive wireless power transfer (IWPT). The goal is a design of IWPT system. The work will be oriented to the modelling of parts of IWPT transmission chain and their realization (especially to structures of induction coils). The optimal properties of the structures will be examined.

Requirements

The student has to know the basics of the electromagnetic field and circuit theory and have basic Matlab skills.
Synthesis of antenna arrays
Ing. Jan Kraček, Ph.D.
1 semester / All students
Department of Electromagnetic Field
11. 7. 2018
The project is aimed at the investigation of antenna arrays. The goal is synthesis of an antenna array according to given requirements, e.g., radiation pattern. The work will be oriented to the development of methods for antenna synthesis (especially orthogonal methods). The realization and measurement of the synthetized antenna array will be also possible.

Requirements

The student has to know the basics of the electromagnetic field and circuit theory and have basic Matlab skills.
Analysis and Assessments in Speech Pathology
Ing. Jan Rusz, Ph.D.
1 semester / Master or doctoral
Department of Circuit Theory
25. 5. 2018
The project is aimed to finding methods that would be based on the analysis of speech pathological audio recordings and would be able to objectively and automatically determine the degree of speech disorders. These methods could be useful for medical experts in determining of speech pathology severity. There are extensive speech pathology databases and MATLAB codes in SAMI group.

Requirements

Basics in the area of Digital Signal Processing, knowledge of MATLAB. Recommended subjects to be studied at the DCT: Biological Signal Processing (PhD subject) Digital signal processing (PhD subject) Phonetic signals and their coding (PhD subject) Basic Signal Processing (available for fall semester) Signal theory (fall semester) Biological signals (spring semester) Digital Signal processing (spring semeste
Analysis and processing of Multichannel Surface EMG
Ing. Jan Sedlák
1 semester / Master or doctoral
Department of Circuit Theory
25. 5. 2018
The project is aimed on processing of multichannel surface electromyographic signals, muscle activity detection and muscle coordination maps. We cooperate with colleagues from the Faculty of Physical Education and Sport, which record surface EMG signals for various sports activities. There are extensives EMG signal databases and MATLAB codes in SAMI group.

Requirements

Basics in the area of Digital Signal Processing, knowledge of MATLAB. Recommended subjects to be studied at the DCT: Biological Signal Processing (PhD subject) Digital signal processing (PhD subject) Phonetic signals and their coding (PhD subject) Basic Signal Processing (available for fall semester) Signal theory (fall semester) Biological signals (spring semester) Digital Signal processing (spring semester)
Spark Plasma Sintering and Plasma Spray Technology of Silicates
Ing. Josef Sedláček, CSc.
1 semester / Doctoral
Department of Electrotechnology
11. 4. 2017
Project is focused on comparison of two technologies, plasma spray coating and newly developed process spark plasma sintering (or field-assisted sintering technology). A set of silicate samples will be used for the comparison. Especially, the micro-structural and electric properties of prepared samples will be considered.
The task is directed to:
1) Theoretical study of processes SPS and plasma spray technology and familiarization with production equipment.
2) Sample fabrication of selected silicates.
3) Metallographic treatment, crystallographic assessment and XRD measurement of samples.
4) Sample preparation for the dielectric measurement.
5) Complex assessment of dielectric and transport properties of samples.

Requirements

Chemistry of silicates, Solid states physics, Physics of dielectrics.
Simulation of electricity consumption/production in remote areas for economic purposes.
Ing. Július Bemš, Ph.D.
1 semester / Master or doctoral
Department of Economics, Management and Humanities
11. 7. 2018
The project is focused on coverage of power consumption by production from renewable sources backed-up by diesel/battery system. The student will simulate electricity consumption and its coverage by various power sources (solar, wind, diesel generator, batteries). The simulation will be based on power flows required for economic calculations (no transient process etc.). The most economically efficient option should be preferred.

Requirements

Requirements: the student should be able to create simulations in Matlab or similar software (or programming language) and have knowledge of electricity networks and finances.
LoRa and Sigfox DoS attack
Ing. Lukáš Vojtěch, Ph.D.
1 semester / Bachelor or master
Department of Telecommunications Engineering
11. 7. 2018
Design and realize experimental attack on LPWAN technology Lora and Sigfox. Take advantage of the experiment and layout of the scene, as in the case of the bachelor thesis "Hauser V., LPWAN / LPN Security in IoT" available at: https://dspace.cvut.cz/handle/10467/64726 (in Czech). Compare the results including statistical data processing.

Requirements

Bases of the measurement techniques of radio transmission
Object indoor localization
Ing. Martin Šipoš, Ph.D.
1 semester / All students
Department of Measurement
12. 4. 2017
Nowadays, the localization of the object outside of buildings can be easily done using GNSS. The situation is more difficult when the localization is performed inside of buildings. The main task of this project is the realization of the system for localization of the object in indoor applications. The system will use cameras, ultra-wideband, lidars, etc. The project cover HW and SW realization, designing of algorithms, image processing.

Requirements

Capability of using MATLAB, C/C++/C# programming, Python, HW realization, image processing.
Quadcopter and camera based tracking
Ing. Martin Šipoš, Ph.D.
1 semester
Department of Measurement
12. 4. 2017
The drones or UAVs (Unmanned Aerial Vehicles) became very popular during last years.
They are used in wide range of applications such as search and rescue, monitoring and mapping of various areas, hobby flights, etc. The UAVs can be remotely controlled or they can flight autonomously. The main task of this project is design and development algorithms that will process data from small-light camera mounted on quadcopter and based on the image processing, algorithmswill control flightof quadcopter.
For example, the quadcopter equipped by camera will follow moving LED diode.
Finally, the algorithms should be implemented and the quadcopter should be in constant altitude and should follow LED, which is below it and which defines the position. As a quadcopter, the small, light Crazyflie or AR.Drone can be used.

Requirements

Matlab programming, C code and Python programming, HW realization, control algorithms.
Deep neural networks for image classification and segmentation
Ing. Michal Reinštein, Ph.D.
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
This project deals with design, implementation and experimental evaluation of state-of-the-art deep neural network(s) and their ensembles when solving real-world problems (public dataset, eventual participation in a Kaggle competition possible) in areas of robotics, autonomous cars, satellite imagery. We shall use tensorflow, python, multi-GPU computations.
Super-resolution for satellite imagery using generative adversial networks
Ing. Michal Reinštein, Ph.D.
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
The aim is to design, implement, and experimentally evaluate deep neural network architecture for solving the problem of increasing resolution (super-resolution) of satellite imagery using Generative Adversial Networks. The contribution will be focused on evaluating the reliability of generated high frequency content that is unobservable in the imagery otherwise. Main tools will be: python and tensorflow.
Continuous testing of CAN bus support in Linux
Ing. Michal Sojka, Ph.D.
semester / All students
Department of Control Engineering
25. 5. 2018
The goal of this project is to build an automated testbed for continuous testing of CAN bus support in Linux kernel. Besides basic functionality tests, we are also interested in performance tests with the aim of finding performance regressions. It is expected that the student will use a PC computer and several boards running embedded Linux to perform the testing and measurement. The outputs will be publicly available to the Linux CAN community.

Requirements

Advanced Linux user with knowledge of C language and scripting.
The use of symbolic execution for testing of safety-related software
Ing. Michal Sojka, Ph.D.
1 semester / All students
Department of Control Engineering
25. 5. 2018
The goal of this project is to examine the use of symbolic execution technique to verify the properties of software used in safety-related (and real-time) systems in automotive application domain. The task will be to use an existing symbolic execution tool such as KLEE to execute the existing control software to discover potential bugs. Based on the results of this experiment, new experiments will be developed with the aim of finding a good strategy of how to effectively find bugs in generic real-time control software.

Requirements

Proficiency and experience in programming, good knowledge of C language.
Using anti-collision system data for radar-like visualization (ST, BT, DT)
Ing. Miroslav Macík, Ph.D.
1 semester
Department of Computer Graphics and Interaction
11. 4. 2017
Small aircrafts or gliders are more and more likely equipped with anti-collision systems. In case of small aircrafts or gliders in EU are these systems typically based on GPS-data broadcasting and use soon-to-be standardized FLARM protocol. Another system that is typical for medium sized aircrafts and airlines is ADS-B. Analyze anti-collision systems and corresponding data protocols. Perform a feasibility study of usage of the provided data for ground based, radar like situation awareness system. Design and implement a prototype system that consists of a prototype FLARM (ADS-B) receiver and corresponding visualization system on a case of a small national airfield. Test the system in real environment and identify pros and cons of such a solution.

Requirements

Basic skills in programming.
Camera-based guidance of a six-axis robot
Ing. Pavel Burget, Ph.D.
1 semester / All students
Department of Control Engineering
25. 5. 2018
An industrial camera is connected to a robotic system, where a PLC and a six-axis robot are connected in a Profinet network, and the image recognition algorithms are part of the camera system. The aim of this one-term project is to program the camera and the robot to search for given objects laying on the ground and have them collected by the robot at one place. This task is going to be extended in such a way that the robot follows the movements of a human in front of itself.
Programming of Profinet diagnostics
Ing. Pavel Burget, Ph.D.
1 semester
Department of Control Engineering
12. 4. 2017
Write communication functions in Ruby to read diagnostics information from Profinet
IO devices. Create filters to get only selected information from the devices. Design
algorithm to analyse the obtained data and provide the status of the devices
(good/bad IO data, diagnostics available, etc.), IO data cycle, etc
Robotic table football
Ing. Pavel Burget, Ph.D.
1 semester
Department of Control Engineering
12. 4. 2017
The aim of this project is to equip a table football with motors and a motion
control system to create a platform for testing of algorithms that allow the
motion control system to play football automatically against a human player.
The outcome of this project is going to be analysis of the movements,
requirements specification and project design description. Additionally the
mechanical platform to be fit on the table football will be created and basic
configuration of the motion control system will be done.
Simulation in digital factory
Ing. Pavel Burget, Ph.D.
1 semester
Department of Control Engineering
12. 4. 2017
Design, create and test simulation of a production line in Tecnomatix digital factory.
The production line is inspired by a welding line in a car factory and contains four
industrial robots with grippers and welding units. There are also conveyors
transporting the products. One of the robot is a real six-axis welding robot. The tools of
two of the robots are changeable. After testing the production line in simulation create
a PLC program to control the line in digital factory and then transfer the robotic
program to the real robot
Analysis and processing of intracranial EEG
Ing. Petr Ježdík, Ph.D.
1 semester
Department of Circuit Theory
12. 4. 2017
Suitable for: Ph.D. and MS. Students

The project is aimed on methods for analyses of intracranial (invasive) EEG in interictal, ictal
and postictal epileptic state in time and frequencies domain. Analyses are focused on seizure onset zone detection,
specific waveform identification and epiletiform-network links finding.
Precise localization and removing of epileptic zones is essential for successful outcome of surgical treatment.
There are extensive EEG signal databases and MATLAB codes in ISARG group cooperating with University Hospital Motol in Prague.

Requirements

Basics in the area of Digital Signal Processing, knowledge of MATLAB Recommended subjects to be studied at the DCT: Biological Signal Processing (PhD subject) Digital signal processing (PhD subject) Phonetic signals and their coding (PhD subject) Basic Signal Processing (available for fall semester) Signal theory (fall semester) Biological signals (spring semester) Digital Signal processing (spring semester)
Fast learning in Bayesian optimization algorithm
Ing. Petr Pošík, Ph.D.
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
Algorithms ECGA (extended compact genetic algorithm) and BOA (Bayesian optimization algorithm) are population-based optimization algorithms. They are among the most powerfull methods for optimization of complex black-box optimization problems with binary representation. Each generation they build a model of the structure of dependencies among individual solution components. The model learning is a time consuming operation. For ECGA, an efficiency enhancement was proposed recently that allows to simplify and accelerate the learning without any negative effect on the algorithm performance. The goal of this project is to implement a similar method of model learning for algorithm BOA, and evaluate the potential positive and negative effects on the algorithm performance.
Fitness Predictors in Genetic Programming
Ing. Petr Pošík, Ph.D.
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
Genetic programming is an evolutionary method for searching for algorithms that solve certain problem, or for searching a structured description of certain system, e.g. in the form of a mathematical expression. Evaluation of a candidate solution is usually done with respect to a big set of test cases which is time consuming. That's why the so-called fitness predictors are used. A fitness predictor is a small subset of the set of test cases which is able to provide the evolution with similar information as the whole testing set. The goal of this project is the exploration, design and evaluation of methods for constructing the fitness predictors.
Linkage Tree Crossover for Differential Evolution
Ing. Petr Pošík, Ph.D.
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
Differential Evolution (DE) is a successfull black-box optimization algorithm for real-valued problems. Its weakness is a class of problems with dependent solution components; the exponential or binomial crossover operators conventionally used in DE are not suitable for them. The goal of this project is to design a new crossover operator based on the linkage tree known from the genetic algorithms with binary representation. The application of this operator is not straightforward and will require a skillful modification.
Multitasking for Symbolic Regression
Ing. Petr Pošík, Ph.D.
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
Evolutionary algorithms are optimization methods applicable to a broad class of problems. However, compared to other, more specific methods, they are slower. One opportunity to increase their speed is to take advantage of their ability to solve several optimization problems at once (multitasking). The goal of this project is to explore this possibility for symbolic regression tasks, i.e. for machine learning problems where the goal is to find a symbolic mathematical expression fitting the available data.
Population size adaptation in evolutionary algorithms
Ing. Petr Pošík, Ph.D.
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
The population size is one of the most important parameters of evolutionary algorithms that affects their performance. There are several existing approaches for adapting the population size during the run of the algorithm (Parameter-less GA, Population pyramid, etc.). The goal of this project is to compare these methods for a chosen type of an evolutionary algorithm and for a chosen class of problems, or to develop your own population size adaptation mechanism and assess its pros and cons.
Portfolio optimization method with success prediction
Ing. Petr Pošík, Ph.D.
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
Design a composite optimization algorithm (portfolio algorithm) which automatically chooses the most promising optimization algorithm in the portfolio to run in the next step. Perform a comparison against its constituent algorithms and against other chosen competitors. The comparison shall reveal the differences during the whole course of optimization, not only after the optimization is finished. The COCO framework is a suitable candidate.
Random embeddings for black-box optimization algorithms
Ing. Petr Pošík, Ph.D.
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
The solution quality of a large-scale optimization problem often depends only on a smaller subset of solution features. Random embedding is a method how to perform the search in a low-dimensional subspace of a high-dimensional search space. The goal of this project is to design a method of using random embeddings in black-box optimization algorithms (local search, evolutionary algorithms): whether the embedding shall be static or dynamic, whether the low dimensionality shall be constant or dynamically changing, how often should they change, etc. The resulting method shall then be compared at least to the chosen optimization algorithm without random embedding.
Restarting optimization algorithms using their predicted performance
Ing. Petr Pošík, Ph.D.
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
In the article below, the researchers optimize the configurations of Deep neural networks, and they throw away bad configurations by extrapolating the learning curves of DNN. The goal of this project is to assess whether and how a similar principle can be applied to black-box search algorithms (local search, evolutionary algorithms, etc.) by observing and predicting their convergence curves. The study can assess various models for algorithm performance prediction, e.g. a parametric exponential model, or a non-parametric model based on resampling the past algorithm improvements.
Adjacent cells in a mobile network
Ing. Robert Bešťák, Ph.D.
1 semester / All students
Department of Telecommunications Engineering
11. 7. 2018
The aim of project is to design and to test algorithms for recognition of adjacent cells in mobile networks.

Requirements

Required skills: working English speaking and writing, basic combinatorial mathematics and statistics Recommended skills: General programming
Models for femtocell networks
Ing. Robert Bešťák, Ph.D.
1 semester
Department of Telecommunications Engineering
12. 4. 2017
The aim of project is to compare different existing models for femtocell placement (random placement /dense-urban model/ etc.) and develop new ones suitable for the femtocell environment. Steps: to familiarize with femtocell problematic –to implement exiting models in Matlab and to analyzetheir features –to design new model(s) suitable for the femtocell environment.

Requirements

Recommended skills: General programming, Matlab
Open source data
Ing. Robert Bešťák, Ph.D.
1 semester / All students
Department of Telecommunications Engineering
11. 7. 2018
Integration of open source data sources to data processing platform. Develop and test interfaces to integrate open source data to Hadoop platform. Automation, quality check, data versions updates.

Requirements

Required skills: working English speaking and writing, basic combinatorial mathematics and statistics Recommended skills: General programming
Use of Ripley's K function in femtocell environments
Ing. Robert Bešťák, Ph.D.
1 semester
Department of Telecommunications Engineering
12. 4. 2017
The aim of project is to investigate how we can describe/ characterize distribution of femtocells in a macrocell, i.e. whether femtocell locations are homogenous, cluster, disperse, etc. To characterize the distribution, Ripley's K functionwill be used.
Steps: to familiarize with femtocellproblematic–to develop a model in Matlab (onemacrocell containing various number of femtocells that are differently distributed in the macrocell) – to characterize the femtocell distribution with the help of Ripley'sK function.

Requirements

Recommended skills: General programming, Matlab
3D reconstruction of historic buildings interiors from a structured light camera
Ing. Tomáš Báča
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
Unmanned Aerial Vehicles (UAVs) are expanding from open outdoor environments into more constrained indoor locations thanks to improvements in accuracy and precision of localization, navigation and control algorithms in recent years. New possibilities for deployment of autonomous UAV swarms into indoor environments emerge, leading to the development of high-level mission-oriented algorithms. Our team is in particular interested in mapping and documenting of historic buildings to assess the condition of the ceiling, murals, statues, stained glass, etc. An accurate and detailed 3D model of the historic buildings is needed for arranging restoration projects, measuring certain dimensions of the interior, and also for planning of further UAV missions. The solution will use structured light approach, which consists of projecting a near-infrared (IR) pattern of parallel lines that are registered by two cameras mounted on a fixed baseline. The depth in the image is estimated from the deformation of the narrow-band pattern. The goal of this project is to autonomously navigate a UAV equipped with a structured light camera through a historic building to capture the interior by laser scans completely. The solution will involve onboard decision making and trajectory replanning that will guide the UAV to yet unmapped areas while simultaneously avoiding collisions with obstacles. When the whole interior is mapped, the dataset obtained during the flight will be processed, and triangular meshes will be built to generate a realistic 3D surface model of the interior.
3D reconstruction of historic buildings interiors from LIDAR scans
Ing. Tomáš Báča
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
Unmanned Aerial Vehicles (UAVs) are expanding from open outdoor environments into more constrained indoor locations thanks to improvements in accuracy and precision of localization, navigation and control algorithms in recent years. New possibilities for deployment of autonomous UAV swarms into indoor environments emerge, leading to the development of high-level mission-oriented algorithms. Our team is in particular interested in mapping and documenting of historic buildings to assess the condition of the ceiling, murals, statues, stained glass, etc. An accurate and detailed 3D model of the historic buildings is needed for arranging restoration projects, measuring certain dimensions of the interior, and also for planning of further UAV missions. The goal of this project is to autonomously navigate a UAV equipped with a rotating laser rangefinder (LIDAR) through a historic building to capture the interior by laser scans completely. The solution will involve onboard decision making and trajectory replanning that will guide the UAV to yet unmapped areas while simultaneously avoiding collisions with obstacles. When the whole interior is mapped, the dataset obtained during the flight will be processed, planar features extracted, and triangular meshes built to generate a realistic 3D surface model of the interior.
Autonomous laser-scan-based relative navigation for UAVs
Ing. Tomáš Báča
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
Unmanned Aerial Vehicles (UAVs) are expanding from open outdoor environments into more constrained indoor locations thanks to improvements in accuracy and precision of localization, navigation and control algorithms in recent years. New possibilities for deployment of autonomous UAV swarms into indoor environments emerge, leading to the development of high-level mission-oriented algorithms. The community achieved significant progress in the development of simultaneous localization and mapping (SLAM) algorithms that estimate the position of the robot in a gradually built map. However, it has turned out, that the position estimate in the map is not necessary for particular tasks, and the position relative to obstacles currently detected by the onboard sensors is more important. One of such missions is the so-called search-and-rescue, in which the UAV has to find a specific object (or person) in an unknown cluttered environment. The goal of this project is to develop a navigation method that accepts a user input consisting of approximate direction of the searched object, and the UAV autonomously flies in the specified direction while modifying its trajectory to avoid obstacles in its vicinity. The area around the UAV is continuously scanned by a rotating laser rangefinder (LIDAR) that generates 360° planar laser scans of the surroundings. The solution should use the most recent laser scan to obtain the nest control input by employing potential field planning or adaptive model predictive control. The mission is accomplished when the searched object is detected by a passive camera or an emergency signal detecting device.
Cooperative mapping of an unknown building by a swarm of UAVs
Ing. Tomáš Báča
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
Unmanned Aerial Vehicles (UAVs) are expanding from open outdoor environments into more constrained indoor locations thanks to improvements in accuracy and precision of localization, navigation and control algorithms in recent years. New possibilities for deployment of autonomous UAV swarms into indoor environments emerge, leading to the development of high-level mission-oriented algorithms. Our team is in particular interested in mapping and documenting of historic buildings to assess the condition of the ceiling, murals, statues, stained glass, etc. The historics and restorers need high-quality detailed images to evaluate the unreachable parts of the buildings for which the UAV is the ideal platform that saves money and time that would otherwise be spent on building scaffoldings to reach the areas of interest. The goal of this project is to develop a technique for documenting the UAV interior using a team of UAVs equipped with cameras. The high quality and resolution of the images will be achieved by stitching multiple close-up images taken by all members of the UAV team. Assuming the approximate positions of the UAVs are known, the camera images taken from a close vicinity of the documented object can be stitched together into a large and detailed image by matching image features. The optimal trajectory for all members of the team has to be planned so that the whole documented object is captured with sufficient overlap between the images to allow stitching into a high-resolution composite image. The time spent on capturing the images has to be minimized since the flight time of the UAVs is constrained by their battery capacity.
Depth estimation in a passive monocular camera image aided by laser scans
Ing. Tomáš Báča
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
Unmanned Aerial Vehicles (UAVs) are expanding from open outdoor environments into more constrained indoor locations thanks to improvements in accuracy and precision of localization, navigation and control algorithms in recent years. New possibilities for deployment of autonomous UAV swarms into indoor environments emerge, leading to the development of high-level mission-oriented algorithms. Our team is in particular interested in mapping and documenting of historic buildings to assess the condition of the ceiling, murals, statues, stained glass, etc. When a UAV is operating in indoor environments, it is vital to prevent collisions of the UAV with obstacles with arbitrary shape. The UAV can be equipped with multiple sensors estimating the distance of an obstacle from the UAV. The sensors are working on different principles (LIDARs and passive cameras) to generate a computer representation of the surroundings of the UAV. While the LIDARs provide an accurate and precise distance to the obstacles, they provide measurements only in the sensor plane. A monocular camera, on the other hand, can detect obstacles in the whole area in front of the UAV, but cannot provide depth information. The goal of this project is to develop a collision avoidance system. The technique will segment the image, to obtain the position estimate of objects in the axes of camera plane. The distance from the UAV to the objects will be estimated by finding correspondences between the objects in the image and laser scans. The trajectory of the UAV will then be modified to avoid the collision with the obstacle.
Development of virtual bumpers for UAVs based ultrasound sensors
Ing. Tomáš Báča
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
Unmanned Aerial Vehicles (UAVs) are expanding from open outdoor environments into more constrained indoor locations thanks to improvements in accuracy and precision of localization, navigation and control algorithms in recent years. New possibilities for deployment of autonomous UAV swarms into indoor environments emerge, leading to the development of high-level mission-oriented algorithms. Our team is in particular interested in mapping and documenting of historic buildings to assess the condition of the ceiling, murals, statues, stained glass, etc. Despite the advanced localization methods, a situation that could lead to a collision of the UAV with a static or dynamic obstacle might occur due to localization algorithm failure, sensor not detecting the obstacle, etc. The goal of this project is to design, build and implement a system that will serve as a low-level, last-resort collision avoidance solution. Such a system sometimes referred to as a virtual bumper, will activate when the safety area defined by a sphere around the UAV is compromised. The UAV would quickly, but smoothly modify its trajectory to avoid the collision by an evasive maneuver with parameters depending on the time to potential collision. The development of the hardware part, consisting of a set of cheap ultrasound sensors mounted on the frame of the UAV, will be a crucial part of the solution.
Mapping of camera images onto 3D surfaces extracted from LIDAR scans
Ing. Tomáš Báča
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
Unmanned Aerial Vehicles (UAVs) are expanding from open outdoor environments into more constrained indoor locations thanks to improvements in accuracy and precision of localization, navigation and control algorithms in recent years. New possibilities for deployment of autonomous UAV swarms into indoor environments emerge, leading to the development of high-level mission-oriented algorithms. Our team is in particular interested in mapping and documenting of historic buildings to assess the condition of the ceiling, murals, statues, stained glass, etc. An accurate and detailed 3D model of the historic buildings is needed for arranging restoration projects, measuring certain dimensions of the interior, and also for planning of further UAV missions. The UAV can be equipped with multiple sensors working on different principles (LIDARs, passive and structured light cameras) that generate a computer representation of the surroundings of the UAV. The goal of this project is to combine the environment representations from LIDAR and passive monocular camera. Features will be extracted in both image and laser scan, and correspondences between them found. The camera images will then be mapped on the surface consisting of triangular meshes constructed from the laser scans.
Simultaneous localization and mapping of a UAV equipped with a LIDAR
Ing. Tomáš Báča
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
Unmanned Aerial Vehicles (UAVs) are expanding from open outdoor environments into more constrained indoor locations thanks to improvements in accuracy and precision of localization, navigation and control algorithms in recent years. New possibilities for deployment of autonomous UAV swarms into indoor environments emerge, leading to the development of high-level mission-oriented algorithms. Our team is in particular interested in mapping and documenting of historic buildings to assess the condition of the ceiling, murals, statues, stained glass, etc. A reliable position estimate of the UAV is needed for all indoor autonomous flights. Since the global navigation satellite system (GNSS) services are generally not available indoor, the UAV must be localized using onboard sensors only. One of the sensors that are widely used for localization of UAVs, is the rotating laser rangefinder (LIDAR). Two principal approaches exist for laser-scan-based localization: scan matching techniques and feature-based methods. The goal of this project will be to develop a feature-based simultaneous localization and mapping (SLAM) system. The system will read laser scans from the LIDAR, process them, extract stable features, find corresponding features in the simultaneously built map to estimate the current position of the UAV in the map. A loop closure detection algorithm will be employed to correct the position drift after returning to a previously visited part of the map.
Simultaneous localization and mapping of a UAV equipped with a structured light camera
Ing. Tomáš Báča
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
Unmanned Aerial Vehicles (UAVs) are expanding from open outdoor environments into more constrained indoor locations thanks to improvements in accuracy and precision of localization, navigation and control algorithms in recent years. New possibilities for deployment of autonomous UAV swarms into indoor environments emerge, leading to the development of high-level mission-oriented algorithms. Our team is in particular interested in mapping and documenting of historic buildings to assess the condition of the ceiling, murals, statues, stained glass, etc. A reliable position estimate of the UAV is needed for all indoor autonomous flights. Since the global navigation satellite system (GNSS) services are generally not available indoor, the UAV must be localized using onboard sensors only. One of the sensors that are widely used for localization of UAVs are the passive stereo cameras. While cameras work well in outdoor environments, their deployments indoor is problematic, due to insufficient light. The solution is to use structured light approach, which consists of projecting a near-infrared (IR) pattern of parallel lines that are registered by two cameras mounted on a fixed baseline. The depth in the image is estimated from the deformation of the narrow-band pattern. The goal of this project is to develop a simultaneous localization and mapping (SLAM) system based on a structured light depth camera. The system will read the depth image from the camera, extract stable features, and find corresponding features in the simultaneously built map to estimate the current position of the UAV in the map. A loop closure detection algorithm will be employed to correct the position drift after returning to a previously visited part of the map.
Economy of energy storage utilization
Ing. Tomáš Králík, Ph.D.
1 semester / Bachelor or master
Department of Economics, Management and Humanities
11. 7. 2018
The aim of the project is to identify and evaluate the feasibility of energy storage utilization under current market conditions. With the rising numbers of intermittent energy sources (renewables) and advancing decommissioning of “classical” sources, new business opportunities for storage are emerging. Therefore this projects is focused on identification of all relevant and technically feasible project, modelling the full life cycle of storage, identification of all potential revenues (and how to fairly distribute them among all participating parties) and also discussion of possible changes of key technical and economical inputs in the near future (based especially on upcoming legislation).

Requirements

Basics in financial management, basics in energy markets, basics in technologies for energy storage, basics in power systems modeling
Adversarial machine learning in adversarial environment
Ing. Tomáš Pevný, Ph.D.
semester / All students
Department of Computer Science and Engineering
25. 5. 2018
Adversarial machine learning study situation, where active adversary aims to fool classifier / detector created by a system trained from a set of samples. We want to study these systems is a relatively simple scenario of detecting abuse of services. We wish to found a boundary on size of the problem which can be solve correctly from game-theoretic perspective, and how violations of optimality influences quality of the solution.

Requirements

Student should be familiary with basic machine learning tools, especially classifiers trained by gradient methods (SVM, logistic regression, kernel methods, and neural networks). Basic knowledge of Game Theory is welcomed
Automated Face Swaping
Ing. Vojtěch Franc, Ph.D.
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
The task will be to develop a system for automated face swapping. The system first learns a model of a target identity from examples of his/her face images. Then, given an arbitrary source face the system creates a new face of the the target identity which preserves pose, facial expression and lighting conditions of the source face.
Learning CNNs from Weakly annotated facial images
Ing. Vojtěch Franc, Ph.D.
1 semester / Master
Department of Cybernetics
11. 7. 2018
Learning Convolutional Neural Networks (CNNs) for face recognition requires large sets of annotated face examples. Manual annotation of faces is laborious work and results are often imprecise. The goal of this project is to implement a system that can learn CNNs from weakly annotated images that can be created by an automated process not requiring humans. As a case study the system will be applied to learn CNN for age and gender recognition from weakly annotated IMDB database that contains 300k faces of movie celebrities. Each image in the database is annotated by the name, biological age and gender of the captured celebrity. However, the images can contain multiple faces and its is not know to which face the annotation should be linked (see example image below). The implemented system should establish the missing link between the images and the annotations.
Framework for benchmarking of motion planning methods
Ing. Vojtěch Vonásek, Ph.D.
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
The task of motion planning is to find a feasible path or trajectory for a robot between two given places in a known environment. Motion planning is studied mainly in robotics and number of methods have been proposed to solve the problem. The goal of this thesis is to create a software framework for automatic testing of motion planning methods. The framework will be accessible using a web interface and it automatically tests desired software. System can be based e.g. on Jenkins

Requirements

Web tools, server configuration and programming in c++ is mandatory.
Management concept of Smart Grids networks
Ing. Zdeněk Brabec, CSc.
1 semester
Department of Telecommunications Engineering
12. 4. 2017
Smart Grids are energy networks, which use information and communication
technologies for real-time management of electricity production and consumption.
Telecommunications is a branch where the network management is mastered very well.
The aim of the work is a complex analysis of existing procedures for managing
telecommunications networks usage for Smart Grids management.
The work will be elaborated in accordance with the activities of the TM Forum organization.
Deep learning for automatic detection of multiple myeloma in CT of femurs
Jan Hering, Dr. rer. nat.
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
The diagnostics of multiple myeloma (a type of bone marrow cancer) uses blood plasma tests as well as CT images of large bones (femurs in this case). From the methodical point of view, these data represent a situation to be found more often in next years -- we don't have fully annotated images at voxel level, since this work is very time consuming, but are provided only with a 'global' label (healthy / diseased), which comes naturally from the diagnosis. By means of Multiple-Instance Learning (MIL) we are able to reach reasonable precision in the classification. However, the interesting question how a deep-learning system will perform is still without answer. The aim of this work is to implement a selected CNN architecture, or to compare multiple architectures, and evaluate it's performance against the current system. This work thus brings the opportunity to learn more about the newest methods in machine learning, to improve own programming skills as well as to get in touch with the field of medical imaging and computer assisted diagnosis.

Requirements

Python
Extensive-form games solution concepts
Mgr. Branislav Bošanský, Ph.D.
1 semester
Department of Computer Science and Engineering
12. 4. 2017
Nash equilibrium (NE) is the best known solution concept in game theory, however, NE does not exactly correspond to the way people make their decisions in competitive scenarios. Therefore, a number of alternative solution concepts was designed, such as quantal response equilibrium, or cognitive hierarchy models. Until now, only a limited focus has been given on studying these solution concepts in the domain of extensive-form games together with the computational aspect. Therefore, the task is to:
1) Study alternative behavioral game-theoretic solution concepts in the domain of extensive-form games.
2) Investigate existing algorithms for finding these solution concepts in extensive-form games.
3) Analyze and design a modification of existing algorithms using an iterative double-oracle approach that is successfully used for computing NE.
4) Implement designed algorithms and experimentally evaluate the performance against human participants.
Transaction costs of selected energy efficiency and RES policy instruments
Michaela Valentová, MSc., Ph.D.
1 semester / Bachelor or master
Department of Economics, Management and Humanities
11. 7. 2018
The negative impact of transaction on implementation of energy efficiency measures has been acknowledged and supported by a number of studies. Transaction costs can hinder implementation of energy efficiency policy measures or even prevent them from being implemented at all. Yet, transaction costs have not been systematically taken into account when designing energy efficiency policies and have not been systematically evaluated ex post, either. Therefore, the information on overall effectiveness of (public) programmes tends to be incomplete. The goal of the work is to analyse a selected energy efficiency or renewable energy policy instrument from the view point of transaction costs and to compare the results with existing studies. The preliminary steps of the task are: 1. Overview of the methods used to analyse transaction costs in energy efficiency or renewable energy policy instruments, 2. Methodology, 3. Analysis of transaction costs in a selected energy efficiency or renewable energy policy instrument, 4. Comparison with existing studies

Requirements

Basics in energy efficiency, basics in renewable energy sources, basics in (energy) economics
Fully Actuated UAVs and their Outdoor Applications
Nicolas Staub, Ph.D.
1 semester / Master
Department of Cybernetics
11. 7. 2018
In the past couple of years the community interest for fully actuated aerial vehicle grew steadily, fostered by projects focused on aerial physical interactions. Nowadays only a handful of labs have working prototypes, mostly indoor. The field is still in its infancy and many research topics are open, in particular outdoor flight with or without physical interactions are particularly promising and unexplored. The domain to be explored comprise, but are not limited, to mechatronic design, planning techniques, sensor fusion, localization, control for physical interaction, estimation of external wrench, external wrench classification and swarming behavior.

Requirements

Nonlinear control theory, estimation techniques, C/C++, Python, ROS, Matlab
Prototype of Outdoor Fully Actuated UAV
Nicolas Staub, Ph.D.
1 semester / Bachelor
Department of Cybernetics
11. 7. 2018
In the past couple of years the community interest for fully actuated aerial vehicle grew steadily, fostered by projects focused on aerial physical interactions. Nowadays only a handful of labs have working prototypes, mostly indoor. The goal of this work is is to participate to the development of a fully actuated prototype able to flight outdoor. In particular, the mechatronic design of the platform will be pursued, merging components from MRS and LAAS-CNRS framework. Integration on the prototype of both hardware and software components will be conducted and validated in both simulation and outdoors experiments.

Requirements

Basics in electronic and mechanic, C/C++, Python, 3D printing, hands on experience
Sensor Fusion for Fully Actuated Outdoor UAVs
Nicolas Staub, Ph.D.
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
In the past couple of years the community interest for fully actuated aerial vehicle grew steadily, fostered by projects focused on aerial physical interactions. Nowadays only a handful of labs have working prototypes, mostly indoor. The goal of this work is to participate to the development a fully actuated prototype able to flight outdoor. In particular, a focus will be given to the estimation of external wrench e.g. to retrieve wind estimate. The development tasks of the prototype will be mostly software oriented (from driver to high-level controller) but can contain mechatronic activities. The work will involve both simulation and outdoor experiments to validate the proposed solutions.

Requirements

Control theory and estimation techniques, filtering techniques, C/C++, Python, ROS, hands on experience preferred
Automatic analysis of gel electrophoresis data
prof. Dr. Ing. Jan Kybic
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
The task is to automatically process images from gel electrophoresis data - suppress noise, detect and localize individual markers, compensate electric current inhomogeneity. Several alternative methods should be tested and results compared. The result should be a standalone program or a web-based application. The method should be applicable to different kind of gels and staining.
Automatic nuchal translucency measurements
prof. Dr. Ing. Jan Kybic
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
Nuchal translucency is a diagnostics measurement performed on fetal ultrasound to determine certain abnormalities. The task is to perform this measurement automatically and reliably, by first assessing the correctness of the pose, identify the correct measurement spot, locally identifying the tissue boundaries and performing the measurement.
Biomedical image analysis competitions
prof. Dr. Ing. Jan Kybic
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
This is a meta-topic. The idea is to choose a medical imaging-related task from either Grand Challenges or Kaggle competitions, solve it and submit a solution. The advantage is that you get an feedback, how well your solution compares to the true state of the art. A motivated team should be able to meet the deadline and actually compete. An individual student might not be able to solve the problem in time to win anything but the comparison should still be possible.
Cartilage and bone segmentation in ultrasound images of the knee
prof. Dr. Ing. Jan Kybic
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
We will work with ultrasound images of the knee for the detection of the osteoarthritis. The task is to segment the cartilage and bone interfaces, extract relevant descriptors, and create a classifier to predict the disease level based on expert evaluations or subjective patient reports. The segmentation may include techniques such as active contours and shape models. Another task involves detecting a classification of osteophytes (bone projections).
Deep learning for 3D drosophila egg segmentation
prof. Dr. Ing. Jan Kybic
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
The task is to detect and segment Drosophila eggs from microscopy images using deep learning (convolutional neural networks). The challenge is that the data are 3D and standard CNN methods cannot be used directly because of memory limitations in available GPU cards. It is therefore necessary to find other methods or reimplement existing ones.
Deep learning for automatic detection of multiple myeloma from CT images
prof. Dr. Ing. Jan Kybic
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
The task is to develop a deep learning (convolutional neural network) method to detect multiple myeloma in 3D CT images of long bones, especially femurs. Several network architectures should be tried and the performance compared with a classical solution. The particularity is that only weak annotations are possible - we know whether a subject is healthy or not but a precise location of the lesion is not available. This leads to so-called multiple instance learning methods.

Requirements

Recommended implementation languages are Python or Julia.
Deep learning for skin cancer classification
prof. Dr. Ing. Jan Kybic
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
Given a skin patch, (i) segment a lesion and (ii) determine, whether it is benign or malignant. Several network architectures should be tried and the performance compared with a standard ABCD method. An additional task is to (iii) create a mobile phone or client/server application for the detection.
Deep learning for tumor detection from histopathological images
prof. Dr. Ing. Jan Kybic
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
The task is to develop a deep learning (convolutional neural network) method to detect cancerous tissue in colorectal histopathological images. Several network architectures should be tried and the performance compared with a classical solution. In some cases, only weak annotations are available - we know whether a subject is healthy or not but a precise location of the lesion is not available. This leads to so-called multiple instance learning methods. Students will be able to compare their results with other authors in the frame of the CAMELYON challenge (https://camelyon17.grand-challenge.org)

Requirements

Recommended implementation languages are Python or Julia.
Deep learning for tumor type classification from histopathological images
prof. Dr. Ing. Jan Kybic
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
The input data are digitized microscopy histology slices of colorectal tissue, with manual or automatic segmentation of the cancerous tissue. The task is to develop a deep learning (convolutional neural network) method to distinguish different tumor types. Several network architectures should be tried and the performance compared with a classical solution.

Requirements

Recommended implementation languages are Python or Julia.
Digital histology microscopy image processing
prof. Dr. Ing. Jan Kybic
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
The input data are digitized microscopy histology slices, from e.g. colorectal or breast tissue. The tasks is to develop image processing algorithms for detection of tissue cells and their nuclei and evaluation of their shape, spatial distribution and other parameters. The next step is to use these parameters (aka features) to classify the tissue into tumor and healthy, and to distinguish between several tumor classes. Students will be able to compare their results with other authors in the frame of the CAMELYON challenge (https://camelyon17.grand-challenge.org)
Efficient finding of nearest neighbors for binary vectors
prof. Dr. Ing. Jan Kybic
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
Binary keypoint descriptors, such as BRIEF or Boosting Binary Keypoint descriptors, are very compact and thus memory efficient. In the image matching task, keypoints are matched using their descriptors. The task is to find an efficient method for finding nearest neighbors for a given query binary vector and to compare it with alternative approaches, e.g. exhaustive comparison or a binary search.
Fast image registration using linear programming and sparse sampling
prof. Dr. Ing. Jan Kybic
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
We shall start from a very fast registration technique [1], which is based on choosing keypoints on class boundaries, sampling their neighborhood in a normal direction, approximating the criterion using piecewise linear functions, and linear programming. Possible tasks include: coupling with a very fast segmentation method, extension to 3D, automatic choice of reliable keypoints, tailor-made innerpoint method optimization, optimized multiresolution and termination criterion, etc.

Requirements

[1] Kybic, Borovec: Fast registration by boundary sampling and linear programming. MICCAI 2018
Fast sparse hierarchical B-spline interpolation
prof. Dr. Ing. Jan Kybic
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
The task of interpolation from a sparse set of points is usually approached by using kriging, i.e. radial basis functions, or a uniform grid of basis functions such as B-splines. To get the best of both worlds, we will attempt to use a sparse set of B-splines are various scales. The non-zero B-splines to be used should be determined automatically. The task is to develop an efficient algorithm for calculating the spline coefficients and to compare this method with alternatives.
Finding objects of known shapes from oversegmentation
prof. Dr. Ing. Jan Kybic
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
The task is to segment an object with clear edges and of approximately known shape, where the shape is described by statistics over some shape descriptors, e.g. moments. We suggest a method which starts by oversegmentating the image into superpixels and then tries to find a compact subset of the superpixels which corresponds to the shape model. We shall try to find efficient algorithms for this task.
From pairwise image registration to sequence image registration
prof. Dr. Ing. Jan Kybic
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
The task is to align a sequence of N images onto a common reference. It is possible to register all images simultaneously, or all possible N^2 pairs but it is very computationally expensive. On the other hand, it is enough to register N-1 pairs and compose the transformations, which is fast but may lead to suboptimal accuracy. We shall (theoretically and practically) explore various ways of choosing the subsets of image pairs to register and how to estimate the individual deformations.
Functional MRI of hypercapnia data
prof. Dr. Ing. Jan Kybic
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
Hypercapnia is a condition of increased presence of carbon dioxide in the blood. Normal reaction of brain to this condition is vessel dilation that will increase the blood flow in the brain in order to compensate for the decreased presence of oxygen in blood. This processes can be monitored over time using a special MRI sequence called BOLD. The BOLD sequence is sensitive to the level of oxygen in blood and increased oxygen concentration shows as increased signal in the images. Hypercapnia BOLD imaging can used to study the vasoreactivy of vessels to the fluctuating level of carbon dioxide in blood. It can be used to identify brain regions with non-standard reaction and as such to identify problematic vessels that may contain stenos or some other vascular pathology. The hypercapnia is usually induced by letting the subjects breath air with increased carbon dioxide concentrations during MRI examination. In our data, we have used several periods of controlled breatholds which is more convenient for use in clinical practice. This, however, brings additional issues in data evaluation as the patients are not always capable to dutifully follow the protocol. The aim of this thesis is to evaluate clinical data from 50+ patients obtained during one or more sessions. The brain of each patient is supposed to contain normal regions and it may contain also pathological regions. The time course of BOLD signal in each patient needs to be examined first to identify what is the normal reaction for each patient (as being close to the expected response). Subsequently, the whole brain of each patient needs to be studied to identify if there are regions with no reaction and with pathological reaction. The result of the automatic analysis of the hypercapnia fMRI data will be compared with clinical findings and the sensitivity and specificity of the automatic method and the BOLD examination will be assessed. Data and clinical findings will be supplied by University hospital in Dresden. The results will be evaluated in close collaboration with medical doctors from the hospital.

Requirements

The aplicant is expected to visit Dresden regularly. The applicant is expected to be fluent in English and have good programming skills in Matlab. Willingness to learn basic principles of fMRI and BOLD imaging as well as to learn the basic medical background of the problem is expected.
Learning multilayer classifiers with weak annotations
prof. Dr. Ing. Jan Kybic
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
This topic is similar to deep learning but without neural networks. We chain several classifiers together - for example the first one will do a pixel-wise segmentations and the second will evaluate the shape of the detected objects. The challenge is to train all classifiers in the chain when training data is only available for the final output - a so-called weak labels. We shall start by simulating several simple cases, gradually moving towards a practical method.
Learning to segment from object counts
prof. Dr. Ing. Jan Kybic
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
The task is to learn to count objects in an image, when the only information available for the training images are the counts. What assumptions are needed to make it possible? We start from simpler cases, e.g. when the segmentation method has a small number of parameters or when pixel-wise annotations are available for at least some of the images and then work towards an efficient and practical method for more difficult cases.
Lung nodule analysis from time series
prof. Dr. Ing. Jan Kybic
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
Lung nodules are tissue masses in the lungs, which may be cancerous. A malignant nodule will grow quickly in time, so it is helpful to compare its image from several time points. A method should be developed to detect nodules, match them across acquisition, and evaluate their growth.
Magnetic particle imaging - using automatically calibrated test targets
prof. Dr. Ing. Jan Kybic
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
Magnetic particle imaging is a recent imaging techniques, which can visualized tiny magnetic particles with high spatial resolution. The bottleneck is the calibration, which is currently performed by moving a point target around the acquisition region. The calibration can be speeded up by using several targets, with known spatial structure. We shall attempt to develop an algorithm which could at the same time identify the spatial structure of the calibration target and the motion of the pattern, simplifying the procedure considerably.
Markov chain Monte Carlo segmentation for fitting geometrical model
prof. Dr. Ing. Jan Kybic
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
Some objects have internal structure - they contain subobjects with a particular geometric relationship. To segment such objects, Markov chain Monte Carlo (MCMC) methods can be used to generate data-driven hypotheses with statistical shape priors. We shall explore methods of this type, first on synthetic data and then to segment real structured objects such as Drosophila eggs, which are known to contain 15 so-called nurse cells. A particular attention will be paid to MCMC efficiency.
Optimal MRI acquisition times for brain tissue segmentation from signal relaxation curve
prof. Dr. Ing. Jan Kybic
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
In MRI, The T1 relaxation time of brain tissue can be measured separately in each voxel from the signal recovery after saturation measured at multiple times. This information can be used to provide brain segmentation with good accuracy without using any prior knowledge about brain topology. What has not been studied yet is the optimal settings of the acquisition times. The influence of more dense sampling and/or presence of noise in the data on the accuracy and systematic error of the segmentation needs to be evaluated. The obtained segmentation will be used for correcting partial volume effect in perfusion MRI measurement. The real data will be provided by the Helmholtz-Zentrum Dresden Rossendorf.

Requirements

The student should learn the basic principles of MR imaging and study the algorithm for brain segmentation from T1 signal recovery, design a simulated experiment and analyze the influence of noise and sampling pattern on accuracy and systematic errors in the segmentation, test the ideal sampling pattern on real data, and optionally demonstrate improvement of the ideal sampling on perfusion MRI.
Piecewise registration for histological slices
prof. Dr. Ing. Jan Kybic
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
When histological specimen is cut into thin slices for subsequent microscopy images, the slice can be broken into several pieces. A specialized image registration algorithm should be developed, which allows very large and uncorrelated displacement of the pieces and much more constrained movement within the pieces.
Registration of ophthalmological sequences
prof. Dr. Ing. Jan Kybic
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
The task is to develop an algorithm for motion compensation of ophthalmological sequences. The task is challenging because the motion is not smooth in time and because there are artifacts due to illumination.
Segmentation and joint-segmentation of aorta from histological and ultrasound images
prof. Dr. Ing. Jan Kybic
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
We have a dataset of ex vivo ultrasound and histological images of the aorta. The final task is to non-invasively detect arteriosclerosis and its prognosis. Current image processing tasks include: 1. Identifying the lumen and different tissue types (including plaques) in the histological images. 2. Identifying the lumen and hopefully also other tissue types in the ultrasound images. 3. Reconstruct 3D volumes and spatially register microscopy and ultrasound data. 4. Joint segmentation of the ultrasound and histological images. 5. Image-based classification of the disease from both the microscopical and ultrasound data.
Surface registration for 3D breast images
prof. Dr. Ing. Jan Kybic
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
A breast surgery is usually performed in a supine position, while preoperative images are acquired in a prone or upright position, hence the need to correct for the resulting deformation. The task is to register preoperative images with a surface scan acquired. A surface registration method needs to be developed, possibly using appropriate mechanical assumptions, such as low compressibility of the breast tissue.
Thickness measurement from ultrasound RF images
prof. Dr. Ing. Jan Kybic
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
For applications such as arteriosclerosis detection, it is interesting to estimate thickness of layers in the blood vessel walls. However, the limited spatial resolution from a standard B-mode image does not allow sufficient accuracy. Processing directly the raw RF signal should enable a better accuracy. This work is a combination of practical experiments with a portable ultrasound machine and implementation of signal processing algorithms.
Ultrasound vessel wall texture classification
prof. Dr. Ing. Jan Kybic
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
Analyze ultrasound images and sequences of carotid arteries to extract the layer thickness, motion and texture of the vessel wall from in-vivo ultrasound images in an attempt to detect and distinguish different types of arteriosclerosis. This is an exploratory work, the challenge is the limited spatial resolution of the ultrasound.
Unsupervised joint segmentation of two images
prof. Dr. Ing. Jan Kybic
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
Given two images of the same scene but from different modalities, the task is to find their common segmentation. This problem is encountered for example in multimodality imaging, or as a subtask of iterative segmentation/registration. One possible formulation is that we want to approximate as well ass possible the target image using a segmentation and an appearance model for each class. We shall study various variations of this idea, e.g. how the approximation quality should be measured and how the resulting optimization should be performed.
Adaptive Optimization
prof. Ing. Filip Železný, Ph.D.
1 semester
Department of Cybernetics
12. 4. 2017
Develop a prototype application of an algorithm that
learns how to solve a certain kind of a combinatorial
problem from examples of small instances and
solutions of that problem, and then applies this
learned knowledge for solving larger instances of the
problem. (Scary as it may seem, this assignment's complexity
may be reduced quite arbitrarily).

Instruction:
The extent of the deliverable will depend on the
particular project category (diploma thesis / Bc. /
semestral project / ...). Agreement will be settled with the supervisor.

Bibliography:
Will be provided by the supervisor.

Realization form:
Code in a language of student's choice + report

Examiner:
Jiří Kléma
Aspects of Conversational Quality Measurements
prof. Ing. Jan Holub, Ph.D.
1 semester / All students
Department of Measurement
25. 5. 2018
Measurement of Conversational Quality as studied by Q15/SQ12 ITU-T is standardized when performed subjectively, e.g. in ITU-T P.800 and ITU-T P.805. However, except of traditional planning algorithm E-model ITU-T G.107, no objective method is standardized or at least widely accepted. Several proprietary algorithms are available, however, some aspects still remain a subject for further study. One of those aspects (echo perception, background noise influence, psophometrical signal level, one-way delay, asymmetrical delay etc.) will be chosen as a topic for the project, short review of available results will be elaborated and proposal for the final algorithm will be given.
Usability and Intelligibility Testing under Physical and Mental Load
prof. Ing. Jan Holub, Ph.D.
1 semester
Department of Measurement
12. 4. 2017
The speech intelligibility tests are well standardized and frequently used for subjective
assessment of telecommunication and radio communication equipment, speech coding
techniques and signal processing algorithms. Recent studies reveal, however, a discrepancy
between laboratory tests following rigid standardized procedures and practical results in field.
The main task of the project will be to design and perform subjective experiment
incorporating suitable (repeatable and well-controllable) mental or physical parallel task,
bringing the test conditions closer to real technology deployment.

Requirements

Ability to work with people, statistical data analysis, off-line data (recorded speech) processing
Interaction of EM Field with Human Body
prof. Ing. Jan Vrba, CSc.
1 semester / All students
Department of Electromagnetic Field
11. 7. 2018
This project is focused on description of interactions of Electromagnetic Fields (EMF) with biological systems (human body above all), mainly from the point of view of physics. Knowledge of these interactions is a basics for research of biological effects of EMF. And it is a gate to prospective use of EMF technology for applications in therapy and/or medical diagnostics. EM simulator Sim4Live will be used for here announced project.

Requirements

The project requires basic knowledge electromagnetics. The ideal student should be interested in EMF simulation (previous experience is not strictly required).
Microwave Thermotherapy in Cancer Treatment
prof. Ing. Jan Vrba, CSc.
1 semester / All students
Department of Electromagnetic Field
11. 7. 2018
Microwave thermotherapy - so called hyperthermia - is one of relatively new cancer treatment methods typically used in combinations with radiotherapy and/or chemotherapy. Main aim of this project will be a study, design and optimization of suitable hyperthermia applicator - i.e. radiating element which distributes EM power into the area to be treated. EM simulator Sim4Live will be used for here announced project.

Requirements

The project requires basic knowledge electromagnetics. The ideal student should be interested in EMF simulation (previous experience is not strictly required).
Energy and economic effectiveness of biomethane production and utilization – fuel cycle approach
prof. Ing. Jaroslav Knápek, CSc.
1 semester / Bachelor or master
Department of Economics, Management and Humanities
25. 5. 2018
Biomethane production is based on anaerobic digestion of biomass or other biodegradable matter (e.g. waste from agriculture) and upgrade of originating biogas to the quality of biomethane (natural gas). Originating biogas can be directly used for power and heat production in cogeneration unit or can be upgraded into biomethane. Biomethane can serve as the substitute of natural gas or other fossil fuels in heat production or can serve as the gaseous fuel for transportation. Biomethane production and utilization can be defined as the series of energy processes starting with biomass planting on agriculture land, through biomass logistics, biomass processing into biogas and biomethane and biomethane delivery to the final customers. The goal of the work is to analyze biomethane fuel cycle (based on intentionally planted biomass on agriculture land – typically fresh maize), to identify individual elements of fuel cycle including their typical technical and economic parameters (for widely used technologies) and to evaluate energy efficiency of given fuel cycle and to evaluate economic competitiveness of produced biomethane against fossil fuels (natural gas, coal) and biogas utilization for power and heat production. Task includes following steps: 1.Discussion of technologies used for biomethane production, collection of data (typical technical and economic parameters) 2.Overview of biomethane production and utilization in EU member states, discussion of possibilities of biomethane utilization and future perspectives of biomethane 3.Analysis of biomethane production fuel cycle, identification of fuel cycle elements, data collection (range of typical energy losses in individual fuel cycle elements, economic parameters) 4.Evaluation of net energy efficiency of biomethane fuel cycle and its comparison with biogas fuel cycle 5.Evaluation of economic effectiveness of biomethane production and its competitiveness against fossil fuels 6.Preparation of economic model of biomethane fuel cycle

Requirements

Requested knowledge: Basics in financial management, basics in renewable energy sources, basics in power and heat production Recommended subjects for the study at FEE: Corporate Finance (AE0B16FIP), Ecology and Economy (AE1M16EKL)
COLLADA Viewer in WebGL
prof. Ing. Jiří Žára, CSc.
1 semester / Master
Department of Computer Graphics and Interaction
11. 4. 2017
Analyze existing software components for reading/parsing COLLADA format for 3D objects/scenes. Implement web-based COLLADA viewer on top of WebGL or other Javascript libraries. The viewer will display arbitrary COLLADA scene and allow camera manipulation using virtual trackball. Implement also the following functionalities: automatic 3D extent calculation, listing of components (triangles, primitives, textures).

Requirements

Basic skills in Javascript programming and computer graphics.
Power System Stability Analysis in Island Operation Mode
prof. Ing. Josef Tlustý, CSc.
1 semester
Department of Electroenergetics
12. 4. 2017
Suitable for: advanced MS. Student

The project aims at the impact of distributed generation (e.g. renewable sources) on operational and control aspects of distribution and transmission systems. The project is focused on the determination of the voltage profile, power losses, power quality and voltage stability. Models of distribution and transmission systems with considerable DG penetration will be used. The influences of DG on voltage stability limits of a power system will be studied. The project case study will be concentrated on island operation modes under different operation conditions.

Requirements

Very good knowledge of power system modeling and power plant operation modes, knowledge of MATLAB and/or other simulation software (Wolfram Mathematica, MathModelica). Recommended subjects to be studied at the Department of Electrical Power Engineering: Electrical Sources and Systems (available for fall semester) Control of Power Systems (spring semester) High Voltage Engineering (spring semester) Power Engineering 2 (spring semester) Power Plants (fall semester) Transmission and Distribution of Electricity (spring semester)
Development of software tool for demonstration of lossy image compression techniques
prof. Ing. Miloš Klíma, CSc.
1 semester / Master
Department of Radioelectronics
11. 7. 2018
The suggested software tool will serve as a demonstration unit for selected image compression technique such as DPCM, VQ, DCT etc. In the introduction, the comprehensive overview of image compression techniques has to be done. After that, the suitable GUI in MATLAB (PYTHON) has to be designed and realized. The GUI has to provide a selection of relevant image compression technique and a definition of required parameters. The implementation of selected image compression techniques is expected. Blocks: Selection of technique and definition of relevant parameters, Compression of selected test picture,Decompression,Demonstration of original, decompressed and difference picture,Calculation of MSE, PSNR or other quality metric.

Requirements

Requirements: knowledge and skills of MATLAB or PYTHON programming
Behavior of the long wire antenna close to the modified conductive ground
prof. Ing. Miloš Mazánek, CSc.
1 semester / Master
Department of Electromagnetic Field
11. 7. 2018
Calculate the radiation pattern of the long wire (non resonant radiation) near to the conductive surface including modified decelerating ground structures. Follow also gain, impedance matching and polarization properties of such antenna. Realize the theoretical study and model, make an experimental structure to verify the radiation pattern. Photo of the structure is available at https://elmag.fel.cvut.cz/projects-in-english

Requirements

Bases of the antenna theory
Theoretical study of the coupling between patches in the patch antenna array (1D) or planar fields (2D)
prof. Ing. Miloš Mazánek, CSc.
1 semester / Master
Department of Electromagnetic Field
11. 7. 2018
Use different motifs of the patch to study and create the model. Minimize coupling, study the influence on gain, impedance, radiation pattern, polarization. Design the structure and measure it. Photo of the structure is available at https://elmag.fel.cvut.cz/projects-in-english

Requirements

Bases of the antenna theory
Gas sensor with surface acoustic wave (SAW)
prof. Ing. Miroslav Husák, CSc.
1 semester / Doctoral
Department of Microelectronics
12. 4. 2017
The project is focused on the design of gas sensor with surface acoustic wave (SAW): The design of the
basic topology of the structure SAW on the piezoelectric material. The design of the SAW sensor model.
The description of the model by the fundamental equations. Creating a 3D model of the structure. The
design of the heating system. Comb electrode structures design on the material surface. Design of SAW
oscillator, determine the operating frequency. Measuring circuit for the characterization of SAW sensors
is based on the method of measuring response of the sensor, which is connected to the oscillator
feedback i.e. positive feedback amplifier, providing an oscillator oscillating at the resonant frequency of
the sensor

Requirements

Basic knowledge and principles of chemical sensors, basic knowledge of electronic circuits, work with ANSYS or Coventor software Recommended subjects to be studied at the DEF: Nanoelectronics and Nanotechnology AE2M34NAN (available for spring semester) Integrated System Structures AE2M34SIS (available for fall semester) Integrated Systems Design AE2M34NIS (available for spring semester) References (the project is closely related to work presented in the following papers): [1] Laposa, M. Husak, J. Kroutil: Gas sensor with SAW structures, proc. of the 2012 NSTI Nanotechnology, Santa Clara, 18.06.2012 - 21.06.2012, pp. 547-577, ISBN 978-1-4665-6275-2 [2] W. Wang, Ch. Lim, K. Lee, S. Yang: Wireless surface acoustic wave chem. sensor for simultaneous meas. of CO2 and humidity, J. of Micro/Nanolithography MEMS MOEMS 8, 3, (2009), 031306. [3] W. Shih, T. Wang, H. Wang, M. Wu: Effect of Alumina Film on Surface Acoustic Wave Properties of ZnO Thin Film Surface Acoustic Wave Devices, Japanese J. of Appl. Phys. 49 (2010) 09MD15. [4] L. Fan, S. Zhang, H. Ge, H. Zhang: „Theoretical optimizations of acoustic wave gas sensors with high conductivity sensitivities“, Sens. and Act. B: Chemical, (2012). [5] R. Ruby: Rev. and comp. of bulk acous. wave FBAR, SMR, IEEE Ultras. Symp., (2007), 1029–40.
Radiation Effects on SiC Power Devices
prof. Ing. Pavel Hazdra, CSc.
1 semester / Doctoral
Department of Microelectronics
12. 4. 2017
Abstract:
The aim of the project is the investigation of the effect of radiation on silicon carbide (SiC)
power devices performance. Advanced SiC power devices (SB or MPS diodes, MOSFETs,
JFETs) will be irradiated with different particles (protons, alphas, electrons or neutrons). The
effect of irradiation on device characteristics will be then investigated for particular projectiles. It
is expected that the applicant will characterize introduced defects, evaluate their influence on
device characteristics and develop adequate models, which allow prediction of SiC device
performance in radiation environment. The project will focus on particular combination of SiC
device and irradiation particle.

Requirements

Requirements: Strong basics in the area semiconductor devices and physics, knowledge of TCAD simulation tools like ATLAS or Synopsys is an advantage Recommended subjects to be studied at the DM: Electronics and Microelectronics AE4B34EM Electron Transport in Semiconductors XP34ETS Modern Power Devices and Integrated Circuits - XP34APD VLSI Structures and Technologies - XP34STV References (the project is closely related to work presented in the following papers): [1] P. Hazdra, V. Záhlava, J. Vobecký, M. Berthou, A. Mihaila, Radiation defects produced in 4H-SiC epilayers by proton and alpha particle irradiation, Materials Science Forum, 740- 742, 2013, pp. 661-664. [2] G. Alfieri, E.V. Monakhov, B.G. Svensson, A. Hallén, Defect energy levels in hydrogen-implanted and electron-irradiated n-type 4H silicon carbide, Journal of Applied Physics, 98, 2005, 113524.
Intelligent detection frame
prof. Ing. Pavel Ripka, CSc.
1 semester
Department of Measurement
12. 4. 2017
Detection frames are used to check persons and their luggage at the entrance to protected areas. New measurement and signal processing methods should be found to increase their feedthrough rate and detectivity. New algorithms should be developer for fusion of magnetic sensors (DC and AC, passive and active), optical cameras, thermal cameras, microwave and ultrasound sensors. These algorithms should be implemented and tested not only on models, but also in real situations. The target of this project is to develop magnetic detection frame with AMR sensors.

Requirements

The project requires knowledge of the basics of electronics, electromagnetic field theory, sensors and instrumentation plus some experience in measurement techniques.
Sensor node for the NDT of building structures
prof. Ing. Pavel Ripka, CSc.
1 semester
Department of Measurement
12. 4. 2017
Failure precaution in civil engineering is fast developing industry. Sensor nodes are placed in the concrete to monitor health of the structure, They placed in the insulation material to monitor humidity and temperature etc.. The goal is the early prevention of possible collapse or deterioration of materials or structures. The aim of the work is to design sensors that can be placed directly into the tested material and operated remotely. The core of the work is to test new types of sensors (e.g. fiber Bragg) of displacement, stress, and humidity. The applicability of the sensors will be verified.

Requirements

Project assumes advanced knowledge in HW/SW and sensor applicability.
Attenuation of shakiness in video sequences
prof. Ing. Pavel Zahradník, CSc.
1 semester / All students
Department of Telecommunications Engineering
11. 7. 2018
The aim of this activity is the development and Matlab implementation of a method for attenuation of shakiness in video sequences, e.g. from drones. This is a 100% software activity.

Requirements

Practical hardware skills are required
Construction of a pulse plasma micro-welder
prof. Ing. Pavel Zahradník, CSc.
semester / All students
Department of Telecommunications Engineering
11. 7. 2018
The aim of this activity is the construction of a pulse plasma micro-welder. This activity comprises 80% hardware and 20% software.

Requirements

Practical hardware skills are required.
Construction of an ultrasonic micro-drilling machine
prof. Ing. Pavel Zahradník, CSc.
1 semester / All students
Department of Telecommunications Engineering
11. 7. 2018
The aim of this activity is the construction of an ultrasonic micro-drilling machine. This activity comprises 95% hardware and 5% software.
Construction of an ultrasonic micro-soldering station
prof. Ing. Pavel Zahradník, CSc.
1 semester / All students
Department of Telecommunications Engineering
11. 7. 2018
The aim of this activity is the construction of an ultrasonic micro-soldering station. This activity comprises 95% hardware and 5% software.

Requirements

Practical hardware skills are required.
Evaluation of a shoting score from a shoting paper target
prof. Ing. Pavel Zahradník, CSc.
1 semester / All students
Department of Telecommunications Engineering
11. 7. 2018
The aim of this activity is the development and Matlab implementation for an automated evaluation of a shoting score acquired using a web cam. This is a 100% software activity in the digital image processing.
Karhunen Loeve and Discrete Cosine Transforms on Digital Signal Processor
prof. Ing. Pavel Zahradník, CSc.
1 semester
Department of Telecommunications Engineering
12. 4. 2017
The aim of this activity is the support of laboratory teaching in the digital signal processing.
This activity comprises programming of both transform on the digital signal processor
(DSP). The programming is supposed in C. The hardware platform is the starter and
development kit based on the floating point digital signal processor TMS320C6748.
This is a 100% software activity.
Laser Beam Deflector
prof. Ing. Pavel Zahradník, CSc.
1 semester
Department of Telecommunications Engineering
12. 4. 2017
The aim of this activity is the design and development of a two-dimensional pointing mirror for the steering of a laser beam.
The control of the mirror is supposed using a fast microcontroller. Its programming is supposed in C. Intermediate hardware and software skills are required.
This activity comprises 50% hardware and 50% software.
Real-time digital signal processing in Matlab
prof. Ing. Pavel Zahradník, CSc.
1 semester / All students
Department of Telecommunications Engineering
11. 7. 2018
The aim of this activity is a real-time imlementation of basic DSP algorithms in Matlab. This is a 100% software activity
Software Radio on Digital Signal Processor
prof. Ing. Pavel Zahradník, CSc.
1 semester
Department of Telecommunications Engineering
12. 4. 2017
The aim of this activity is the support of laboratory education in the digital signal processing.
This activity comprises programming of known algorithms for a direct conversion radio receiver on the digital signal processor (DSP) including filtering and demodulation.
The programming is supposed in C.
The hardware platform is the starter and development
kit based on the floating point digital signal processor TMS320C6748.
This activity comprises 20% hardware and 80% software.
Support for Laboratory Education in Digital Signal Processing in Matlab
prof. Ing. Pavel Zahradník, CSc.
1 semester
Department of Telecommunications Engineering
12. 4. 2017
The aim ofthis activity is the support of laboratory education in the digital signal processing.
This activity comprises the preparation of selected laboratory tasks in the Matlab environment.
It is focused on processing of audio signals in real time using recent Matlab options.
The signal processing includes signal filtering and signal analysis.
This is a 100% software activity.
Diversity techniques for Wireless Optical Networks
prof. Ing. Stanislav Zvánovec, Ph.D.
1 semester / Doctoral
Department of Electromagnetic Field
11. 4. 2017
The project is aimed on analyses and propagation modeling of single and multiple free space optical (FSO) links influenced by atmospheric phenomena. Along with theoretical part, measurements related to channel characterization will be carried out in laboratories of Czech Technical University in Prague and data from outdoor FSO network will be analyzed. The challenging study will involve deep analysis of topological properties of FSO networks and their implementation within ad-hoc networks.

Requirements

Strong basics in the area of electromagnetics and optics, knowledge of MATLAB.
Camera Phone Image Quality
prof. Mgr. Petr Páta, Ph.D.
1 semester / Bachelor or master
Department of Radioelectronics
11. 7. 2018
Mobile phone cameras are popular photographic tools for many users. These cameras very often replace digital cameras. Unfortunately, there is no detail information about imaging quality of these devices. The student will focus on preparing automatic tool in MATLAB for imaging quality assessment based on the standard IEEE 1858. Used metrics should contain spatial frequency response, local geometric distortion, and colour uniformity. More details: IEEE Standard for Camera Phone Image Quality," in IEEE Std 1858-2016 (Incorporating IEEE Std 1858-2016/Cor 1-2017) , vol., no., pp.1-146, May 5 2017, doi: 10.1109/IEEESTD.2017.7921676

Requirements

Signal processing course, Calculus, linear algebra, image processing, MATLAB
Hyperspectral image compression
prof. Mgr. Petr Páta, Ph.D.
1 semester / Master or doctoral
Department of Radioelectronics
11. 7. 2018
Hyperspectral imaging is a modern technique used in many application (natural science, remote sensing, etc.). One of the crucial tasks is compression of acquired images and preparing for the processing. The project is focused on this part. The first step will be an introduction to state of the art of the compression algorithm of hyperspectral images and analyse of the statistical parameters acquired images. The main part of work will be devoted to design novel coder parameters based on PCA.

Requirements

Signal processing course, Calculus, linear algebra, image processing, MATLAB
Preservers in Quantum Information Theory
prof. RNDr. Jan Hamhalter, CSc.
1 semester
Department of Mathematics
11. 4. 2017
Recently many new results on preservers of important quantities in quantum information theory such as mutual quantum information and entropy have appeared. However, many open and challenging questions in this area remain. In the project proposed we would like to concentrate on characterizing maps on quantum systems preserving important information metrics. The method will be based on studying existing methods in the literature and own investigation of the properties of quantum information for finite dimensional spaces. The work should result in a treatment and possibly in journal publication.

Requirements

Strong background in Linear Algebra and Calculus.
Web Graph Service - online graph exploration tool
RNDr. Marko Genyk-Berezovskyj
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
The first version of an integrated on-line tool for exploration of simple undirected graph is available at graphs.felk.cvut.cz. It includes the expandable database of graphs and the graph properties calculator. The task is to extend the project in one or more directions described at http://cmp.felk.cvut.cz/~berezovs/wgs/motiv.html (in Czech).
Rolling shutter benchmark
RNDr. Zuzana Kúkelová, Ph.D.
1 semester / Bachelor or master
Department of Cybernetics
11. 7. 2018
CMOS sensors that are used in vast majority of today’s consumer cameras, smartphones etc. use the rolling shutter (RS) mechanism to capture images. The key difference is that with the global shutter, the entire image is exposed to the light at once, whereas when using the RS the individual image rows (or columns) are captured at different times. When a RS camera moves while capturing the image, several types of distortion such as smear, skew or wobble appear. A perspective camera model is no longer valid in this case and that can be a problem when using methods assuming this model. Recently several algorithms for calibrating RS cameras have been proposed. These algorithms are usually evaluated on a very limited set of images (videos) without a ground truth. While there exist many benchmark datasets with ground truth camera calibrations and positions for perspective cameras, such datasets for RS cameras are still missing. The goal of this project is to create a benchmark dataset for RS cameras. The student/students will collect images/videos for different camera setups (a single RS camera, a stereo rig, two cameras with different rolling shutter directions or different frame rates, different types of camera movements (translations, rotations) during the image exposure….). Ground truth calibrations will be obtained using standard calibration methods, using a global shutter camera, and a controlled motion. For bachelor/master thesis the student will evaluate different state-of-the-art methods for RS calibration on the new proposed dataset. Based on the obtained results, the student will try to address challenging configurations (degenerate configurations) for different RS models.

Requirements

Matlab, basic knowledge of calibration procedures