Popis předmětu - BE4M33TDV

Přehled studia | Přehled oborů | Všechny skupiny předmětů | Všechny předměty | Seznam rolí | Vysvětlivky               Návod
BE4M33TDV Three-dimensional Computer Vision Rozsah výuky:2P+2C
Garanti:Šára R. Role:PO Jazyk výuky:EN
Vyučující:Šára R. Zakončení:Z,ZK
Zodpovědná katedra:13133 Kreditů:6 Semestr:Z

Anotace:

This course introduces methods and algorithms for 3D geometric scene reconstruction from images. The student will understand these methods and their essence well enough to be able to build variants of simple systems for reconstruction of 3D objects from a set of images or video, for inserting virtual objects to video-signal source, or for computing ego-motion trajectory from a sequence of images. The labs will be hands-on, the student will be gradually building a small functional 3D scene reconstruction system and using it to compute a virtual 3D model of an object of his/her choice.

Cíle studia:

To master conceptual and practical knowledge of the basic methods in 3D computer vision.

Osnovy přednášek:

1. 3D computer vision, its goals and applications, course overview.
2. Geometry of points and lines in plane, ideal points and lines, point and line representations, line intersection and point join, homography.
3. Perspective camera model, center of projection, principal point, optical axis, optical ray and optical plane. Vanishing point and vanishing line, cross-ratio of four colinear points and its use for camera calibration. Radial distortion models.
4. Camera resection from six points, external camera orientation from three points.
5. Epipolar geometry, its representation by the fundamental matrix, essential matrix and its decomposition.
6. The seven-point algorithm for fundamental matrix estimation and the five-point algorithm for essential matrix estimation. Triangulation of points in 3D space from image correspondences.
7. The concept of algebraic and reprojection errors and Sampson approximation for reprojection error. Sampson error for fundamental matrix estimation.
8. Local optimization of Sampson error, derivation of a robust error by marginalization of a probabilistic model.
9. Robust optimization of geometric problems in 3D vision by MCMC methods.
10. Reconstruction of a multi-camera system, the bundle adjustment method, minimal and non-minimal representations for some basic geometric objects and mappings on them.
11. Introduction to stereoscopic vision, epipolar rectification of images.
12. The uniqueness, occlusion, ordering, coherence and continuity constraints in stereo and their representation in stereoscopic matching table.
13. Marroquin's greedy algorithm and the maximum-likelihood algorithm for stereoscopic matching.
14. Photometric stereo, its calibrated and uncalibrated versions.

Osnovy cvičení:

1. Introduction, term project specification, instructions on how to select an object suitable for 3D reconstruction, on image capture, and on camera calibration.
2. An introductory computer programming exercise with points and lines in a plane.
3. An exercise on the geometric description of perspective camera. Robust maximum likelihood estimation of a planar line.
4. Computing sparse correspondences by WBS matcher.
5. A computer exercise with matching and estimation of two homographies in an image pair.
6. Calibration of poses of a set of cameras.
7. Midterm test.
8. Sparse point cloud reconstruction.
9. Optimization of point and camera estimates by bundle adjustment.
10. Epipolar rectification and dense stereomatching. Dense point cloud reconstruction.
11. 3D surface reconstruction.
12. Presentation and submission of resulting models.

Literatura:

R. Hartley and A. Zisserman. Multiple View Geometry. 2nd ed. Cambridge University Press 2003.

Požadavky:

Znalost ekvivalentní předmětům Geometrie počítačového vidění a grafiky a Metody počítačového vidění. Detailní aktuální informace o běžícím předmětu na http://cw.felk.cvut.cz/doku.php/courses/a4m33tdv/start

Webová stránka:

https://cw.fel.cvut.cz/wiki/courses/tdv/start

Klíčová slova:

computer vision, digital image and video processing

Předmět je zahrnut do těchto studijních plánů:

Plán Obor Role Dop. semestr
MEOI5_2018 Počítačové vidění a digitální obraz PO 3
MEOI5_2016 Počítačové vidění a digitální obraz PO 3


Stránka vytvořena 9.12.2019 17:50:40, semestry: Z,L/2020-1, L/2018-9, Z,L/2019-20, připomínky k informační náplni zasílejte správci studijních plánů Návrh a realizace: I. Halaška (K336), J. Novák (K336)
Za obsah odpovídá: doc. Ing. Ivan Jelínek, CSc.