Through eye tracking technology, scientists have explored the eyes diverse aspects and capabilities. There are many potential applications that benefit from eye tracking. Each benefit from advances in computer technology as this results in improved quality and decreased costs for eye-tracking systems.This thesis presents a computer vision-based eye tracking system for human computer interaction. The eye tracking system allows the user to indicate a region of interest in a large data space and to magnify that area, without using traditional pointer devices. Presented is an iris tracking algorithm adapted from Camshift; an algorithm originally designed for face or hand tracking. Although the iris is much smaller and highly dynamic. the modified Camshift algorithm efficiently tracks the iris in real-time. Also presented is a method to map the iris centroid, in video coordinates to screen coordinates; and two novel calibration techniques, four point and one-point calibration. Results presented show that the accuracy for the proposed one-point calibration technique exceeds the accuracy obtained from calibrating with four points. The innovation behind the one-point calibration comes from using observed eye scanning behaviour to constrain the calibration process. Lastly, the thesis proposes a non-linear visualisation as an eye-tracking application, along with an implementation.

In the modern day, recognition and tracking of face or the iris is potentially one of the most powerful ways of differentiating between an authentic person and an imposter. Our method uses stereo vision to track the 3-Dimensional coordinates of a target equivalent to a person’s eyes and using a pan-tilt unit we target these areas for additional processing such as iris or facial imaging. One of the most important parts involved in tracking is the way the pan-tilt unit is calibrated. There have been techniques in the past where PTZ (Pan-tilt-zoom) digital camera has been used and calibrated using self calibration techniques involving a checker board calibration grid but the tracking error was found to be large in these techniques. We introduce a more accurate form of calibration of the pantilt unit using photogrammetric calibration technique and view the pan-tilt unit as an emulation of a Pinhole Lens Model to detect and track the target. The system is demonstrated on ideal targets.

Made available in DSpace on 2016-03-15T19:37:51Z (GMT). No. of bitstreams: 1 Andre Brasiliano da Silva.pdf: 4815626 bytes, checksum: e53fa837ff6a7eb3cff0f55f4b3b26ac (MD5) Previous issue date: 2014-08-21 / Eye tracking is an important issue for Human Computer interaction, mainly for users with hand-eye coordination problems. The work presented here shows a low cost and robust eye tracking system capable to work with an HD stream. The implementations used in this work over the base system present diferent techniques in all stages, from face detection to iris detection. Local processing is used in most stages in this implementation, delimiting the region of interest (ROI) for face detection, eye detection and iris detection. The system robustness allow the eye tracking system to control the mouse using eye movements allowing disable users to communicate through a communication interface. The hardware required is simple and based in an high definition webcam. The face detection and eye detection processes are based on the Viola Jones technique; iris detection and tracking are based on the Hough Transform. The usage of local processing reduces the computational cost and even working with high definition stream leads to a performance 33% better than the base system. The system presented here was compared with a commercial system and a set of equipment were tested in order to dene the best set up for the eye tracking system and to validate the work presented here. Future work is presented at the end in order to allow the project continuity. / O rastreamento ocular para usuários com problemas motores é um estudo importante na área de Interface Humano-Computador (IHC). Com o objetivo de fornecer um sistema de rastreamento ocular de baixo custo, este trabalho apresenta uma nova abordagem para um sistema robusto e com alto desempenho. Com relação ao trabalho base para esta pesquisa, a implementação proposta contém inovações em todas as etapas do processo envolvendo o rastreamento ocular, desde a detecção da região da face e dos olhos até a detecção da íris. Neste trabalho, foi utilizado o conceito de processamento local, delimitando as regiões de interesse em todas as etapas do processo: detecção da região da face, região dos olhos e região da íris. Este trabalho permite que pessoas possam efetuar ações controlando o mouse através do movimento dos olhos em uma interface de rastreamento ocular, utilizando apenas equipamentos de uso comum, como, por exemplo, uma webcam. O processo de detecção da face e detecção ocular foi feito através da técnica de Viola e Jones. Para a detecção e rastreamento da íris foi utilizada a Transformada de Hough, e utilização de regiões de interesse com o objetivo de limitar a área de processamento da imagem, e consequentemente, o custo computacional, resultando em uma aplicação com um melhor desempenho e robustez em todas as etapas. Obteve-se um ganho de até 33% em relação ao tempo de processamento do sistema, quando comparado com o sistema base, porém, operando com imagens em alta definição. Foi realizada ainda uma comparação com sistemas de rastreamento ocular de uso comercial e diferentes tipos de equipamentos para validar as técnicas estudadas neste trabalho.