Module 06-25024 (2018)
|Hyung Jin Chang||Semester 2||20 credits|
Vision is one of the major senses that enables humans to act (and interact) in (ever)changing environments. In a similar vein, computer vision should play an equally important role in relation to intelligent robotics. This module will focus on the fundamental computational principles that enable to convert an array of picture elements into structural and semantic entities necessary to accomplish various perceptual tasks. In a series of lectures, we will study the problems of low level image processing, recognition, categorisation, stereo vision, motion analysis, tracking and active vision. The lectures will be accompanied by a series of laboratory exercises where many of these computational models will be designed, implemented and tested in real-world scenarios.
The aims of this module are to:
- give an appreciation of the issues that arise when designing computational models that convert visual signals to structural and symbolic descriptions
- provide an understanding of the state-of-the-art methods and techniques for processing visual information
- give hands on experience of designing, implementing and testing computer vision algorithms in realistic scenarios
- encourage independent thought on deep scientific issues related to visual cognition
On successful completion of this module, the student should be able to:
- design, implement and test simple computer vision algorithms
- write a detailed report on a computer vision project
- survey and critically discuss the research literature in one subfield of computer vision
- demonstrate an understanding of the main computer vision methods and computational models
2 hrs lectures per week, 4 student presentations, laboratory sessions
Contact Hours: 44
Sessional: 2 hour examination (70%), Continuous assessment (team project) (30%).
Supplementary (where allowed): Examination (70%) (with 30% CA carried over).
* Why computer/robot vision * Applications * Computer and human vision perspectives * Challenges 2. Image formation, low level image processing * Image acquisition * Noise removal (Linear filters, Median filter) * Edge detection 3. Structure extraction * Parametric fitting * Hough transform * RANSAC 4. Segmentation * Clustering * K-means * Mean-shift * Graph-cuts 5. Local features * Interest points * Harris detector, Hessian detector * SIFT 6. 3D reconstruction * Stereo vision * Correspondence * Epipolar geometry 7. Recognition * Histograms * Subspace representations * Principal component analysis 8. Categorization * Bag-of-features * Part-based methods * Deformable part-based detector * Hierarchical compositional architectures 9. Motion and tracking * Optical flow * Tracking as detection * Kalman filter 10. Active vision * Perception-action cycle * Attention * Visual servoing