Machine Perception

Subject description

Lectures:

Overview of the field of Machine perception and scientific challenges
Image processing
1. Image formation
2. Binarization, morfology, segmentation
3. Colour spaces and colour perception
4. Linear and nonlinear filters
Image derivatives and edge perception
1. Derivative-based edge perception
2. Edge-based object perception
3. Parametric shape perception
Model fitting
1. Normal equations
2. Homogenous systems
3. Robust approaches
Local features
1. Corner perception
2. Local descriptors in scale space and affine adaptation
Stereoscopy and depth perception
1. Calibrated and uncalibrated systems and reconstruction
Object recognition
1. Subspace methods (PCA, LDA)
2. Local-features-based recognition
Object detection
1. Visual features and detection approaches
Motion perception
1. Local motion perception and object tracking

Exercises:

Exercises will take a form of project-oriented exercises in properly equipped student laboratories. Students will implement various algorithms and test them on different datasets using a variety of sensor systems. Exercises will support an in-depth understanding of the theory. They will also encourage independent thinking and creativity.

The subject is taught in programs

Objectives and competences

In the framework of this course, the students will acquire concrete knowledge and skills in the area of machine perception. The students will develop competences in low-level image processing, 3D geometry of stereo systems, object detection, object recognition, and motion extraction in video sequences. The students will also practice mathematical basics crucial for solving demanding engineering problems, which are essential for analysis of complex signals such as images and video.

In addition, the students will obtain the following competences:

The ability to understand and solve professional challenges in computer and information science.
The ability of professional communication in the native language as well as a foreign language.
The ability to independently perform both less demanding and complex engineering and organisational tasks in certain narrow areas and independently solve specific well-defined tasks in computer and information science.

Teaching and learning methods

Lectures, laboratory exercises in computer classroom with active participation. Individual work on exercises. Theory from the lectures made concrete with hands-on laboratory exercises. Special emphasis will be put on continuous assessment at exercises.

Expected study results

After completing this course a students will be able to:

– understand computer technology and computational methodology for use and development of components for machine vision systems,

– understand the basics of low-level image processing,

– understand the basics of 3D geometry of stereo systems,

– understand the basics of object detection, object recognition,

– know basic motion extraction techniques in video sequences,

– analyze modern computer vision and machine vision algorithms,

– use computer technology and computational methodology for specific applications of autonomous intelligent cognitive systems.

Basic sources and literature

Obvezna:

D. Forsyth and J. Ponce, Computer Vision: A modern approach, Prentice Hall 2011.
R. Szeliski,Computer Vision: Algorithms and Applications, Springer, 2011

Dopolnilna:

H. R. Schiffman: Sensation and Perception, An Integrated Approach, John Wilez & Sons 2001.

Izbrani članki iz revij IEEE PAMI, CVIU, IJCV, Pattern Recognition (dostopno na spletu)