Estudia

• Apply techniques of modelling and calibrating computer vision systems.
• Compute 3D information of the real world from 2D image projections
• Apply the principles of stereovision and triangulation
• Understand the limitations of some feature detecting and feature matching algorithms and how to remove false data associations
• Oral expression
• Team work
• Building creative proposals

1. Introduction

          1.1. Course organization: Objectives, Overview, Contents, Bibliography, Evaluation, Practical Sessions

2. Camera Modelling and Calibration

          2.1. The pinhole camera

          2.2. Intrinsic and extrinsic parameters

          2.3. Computing the calibration matrix

          2.4. Accuracy Evaluation

3. Image Primitives

          3.1. Interest point detectors

          3.2. Harris and Hessian detectors

          3.3. Similarity measures: SAD, SSD, Correlation

          3.4. Introduction to Scale invariant features

4. Correspondence and Planar Transformations

          4.1. Review of SIFT

          4.2. A hierarchy of transformations: Euclidean, Similarity, Affine, Projective

5. Outlier Rejection

          5.1. Probabilistic methods

          5.2. Computing the homography matrix

          5.3. Random Sampling Consensus

          5.4. Applications: Planar motion estimation, Mosaicing, etc.

6. Reconstruction from 2 views

          6.1. The principle of Triangulation

          6.2. Epipolar geometry

          6.3. Computing the Fundamental matrix

          6.4. Accuracy Evaluation

          6.5. Experimental Results

7. Cameras in the real world

          7.1. Comercial cameras

          7.2. Camera characteristics

• Hartley, Richard (2003). Multiple view geometry in computer vision (2nd ed.). Cambridge [etc.]: Cambridge University Press.
• Ma, Yi (2004). An Invitation to 3-D vision : from images to geometric models. New York [etc.]: Springer, cop..

Activitats d'avaluació:

15% Practical Sessions: Lab on corner detection
40% Practical Sessions: Lab on Invariant features
15% Practical Sessions: Lab on 3D laser
(Every lab requires a mark >= 5/10)

30% Final exam
(the grade of the exam should be >=4/10)

To pass the module, the global mark must be >= 5/10

Criteris específics de la nota «No Presentat»:
Lab assignments are mandatory. Failure to deliver a lab assignment implies that the student will not be evaluated in the module.