Self-Locating by Pattern Tracking
This work is my second project of Computer Vision Group, Brno University of Technology that I have solved. Aim of this project is to develope system, which is able to determine position a orientation of camera itself by watching designed pattern. Camera see only small part of this pattern.
You can see my another computer vision projects:
Leucocyte Segmentation and Classification
Self-Locating by Pattern Tracking
Basic principles
System must be very simple to be able work in real time on basic computer, what was requested. Each picture of video stream with 20 FPS speed is processed separately and both position and orientation is computed from that one image.
At first, straight lines are detected on image. Beside the rotation of lines, we have to determine also direction of line, which is important for further calculations. Each line is followed by circular marks at specified side, so by it is position (if it is left or right to line) we can determine line orientation.
After this procedure, the best visible circular mark is detected and number of dots on it is counted. Then, relative position of camera view center according to this mark is computed. When we know the rotation of camera and relative position of mark and center of the view, we can determine absolute position in 2D space, what is desired result.
I reccomend you to see video presentation below, which is the best way to get informations about this project.
Video presentation
Documentation
Here you can find some papers about solving out this interesting task. Documentation is available only in Czech language.
size: 481.87kB, last modified: 19.08.2014