Development of a hybrid algorithm for real-time pupil tracking system

This thesis focuses on eye tracking algorithms that facilitate computer access by using the power and sensitivity of looking interaction, especially for users with limited motor skills due to ALS, spinal cord injury and other disorders, who have low mobility or do not interact with their environment. The fact that the current eye tracking systems need external infrared cameras and create additional costs, the motivation of this study has been determined as an algorithm that will work on the cameras built into the portable computers, and a new algorithm has been developed in addition to the existing algorithms. In this study, a new eye tracking algorithm has been tried by creating a hybrid model VJVK (Viola Jones and Correlation) between the correlation and the Viola Jones algorithm, which allows the detection and matching of objects of unknown size, working with high accuracy and efficiency, for the detection of a face and eye region. In order to observe the difference between Viola-Jones and VJVK, face detection and eye region detection studies were carried out for the two models, and they were interpreted in terms of performance by experimenting with different light and distances. The difference between Viola-Jones and VJVK was compared in percentage terms in terms of speed and accuracy, and a system capable of real-time pupil tracking was achieved with VJVK.