Photogrammetric Computer Vision

The course gives an introduction to the basic concepts of sensor orientation and 3D reconstruction. The goal is an understanding of the principles, methods and applications of image-based measurement. It covers topics such as the algebraic projective geometry, imaging geometry, calibration, orientation methods, stereo image matching and further surface reconstruction methods.

Please notice: The materials for our lectures and exercises are only available through the network of the Bauhaus-Universität Weimar.

Lecture

Introduction

  1. Administration, image-based 3D reconstruction, historical notes, classification of photogrammetry, field of application (slides, print)

Projective 2D geometry

  1. General mathematics, projective geometry, homogeneous 2D coordinates, points and lines (slides, print)
  2. Normalization and interpretation, matrix algebra, planar transformations, concatenation (slides, print)
  3. Estimation of projectivity, 2D homography computation, linear homogeneous equation systems, planar rectification (slides, print)

Projective 3D geometry

  1. Points and planes, spatial transformations, geometric camera model (slides, print)
  2. Projection matrix, interpretation, spatial resection/DLT, optical imaging with lenses (slides, print)
  3. Lens distortion, imaging techniques, ...

Projective multi-view geometry

  1. Relative orientation, epipolar geometry (slides, print)
  2. Fundamental matrix, spatial intersection, projective reconstruction (slides, print)
  3. Essential matrix, normal case, trifocal geometry, trifocal and multi-view tensors (slides, print)
  4. Tensor notation, bundle adjustment (slides, print)
  5. Euclidean and metric upgrade, auto-calibration, ... 

Stereo image matching

  1. Robust parameter estimation, M-estimators, RANSAC (slides, print)
  2. Correspondence problem, image matching constraints, area-based image matching, similarity measures (slides, print)
  3. Least-squares matching, global matching strategies, normal image generation (slides, print)
  4. Summary (slides, print)

Exercise

The exercises will take place biweekly on Wednesdays (13:30 - 15:00)  in Lecture Hall 3 (Hörsaal 3) in Coudraystraße 13b. For relevant study materials, exercise sheets and assignment submissions, please visit the Moodle page for Photogrammetric Computer Vision. An enrollment password will be announced during the first exercise on 15. October 2018.

Over the course of this semester, the following topics will be covered:

1. Exercise: Points and lines in the plane, first steps in MATLAB

2. Exercise: Projective transformation (Homography)

3. Exercise: Camera calibration using direct linear transformation (DLT)

4. Exercise: Orientation of an image pair

5. Exercise: Projective and direct Euclidean reconstruction

6. Exercise: Stereo image matching

Exam

Written examination

Date: Monday, February 11, 2019 at 11:00 am in Audimax, S6 F
Auxiliary resources: none

Preparation material