Detekce očí a jejich jednotlivých částí ve videosekvencích

Abstract

The thesis is focused on designing the methods and creating the software for detecting different parts of eyes in the images taken by a camera. This is done by finding the specific area of eye, which is followed by detecting the iris, locating the pupil, and identifying the eye lids. The system works with the grayscale images that can be taken using both natural and infrared light. My primary motivation for choosing this topic was the anticipated use of this detection system for monitoring the drivers and measuring their drowsiness while they are driving. This should help to increase the safety on roads and motorways. In my work, I describe the methods that are already known, as well as the methods I have modified, and the methods that were created by myself. The final practical outcome of this diploma thesis is the implementation of the methods that can recognize various parts of eye. The implementation is done in C++. The work also includes a detailed description of the techniques that were used. It also contains a thorough experimental evaluation including efficiency and time complexity, which was evaluated for over 1500 images.