In my thesis I’ve done research on how to create an inexpensive gaze tracking system based on image processing methods. There are several fields of application where a gaze tracking system can be used. First I studied the two main fields — research and practical — to show the potential of an efficient and robust tracking solution.
Then I evaluated the Hough-transform, the Viola–Jones cascade classifier and blob-based methods. My objective was to find out which one is the best fit for solving the problem of pupil tracking — the cornerstone of this area. From the point of efficiency and robustness, I found the blob-based tracking to be the most reliable method for implementing the pupil tracking.
Based on blob analysis I’ve presented a solution for a working gaze tracking system using the OpenCV library for image processing, and the Qt framework for building the application’s user interface. The implementation was done using the standard C++ programming language and using the above mentioned libraries a robust and cross-platform tracking solution can be made.