Detection and tracking of open eyes in video sequences
Tutor / director / evaluatorTarrés Ruiz, Francisco
Document typeBachelor thesis
Rights accessRestricted access - author's decision
The objective of this project is to detect eyes on the video footage and propose new applications such as controlled by the eye tracking or at an international security level computers, replace passports for this type of biometric identification (currently used the fingerprints and facial recognition). This project works with images of faces and eyes, taking into account the following: the resolution and image size, because depending on the image resolution methods tested detection will be more or less efficient and if lower to a certain size detection will not work. It has taken into account the lighting conditions of the photo as well. The first part of the project consists of the description of recognition methods that have been chosen, these are: Gabor filters, Circular Hough Transform algorithm and Viola-Jones applied to the detection of an eye, all detection methods modified for detection of objects eyes. Following the implementation of these methods is explained, all developed in Matlab. We used two databases to obtain the results are presented below. The first database used was the "GTAV FaceDataBase" and is a DDBB of 379 images of faces in which you can find faces of men and women, looking straight ahead without making any gesture, looking forward gesturing, covering a portion of the face, profile pictures of faces and head bowed 45 degrees to the left and right. The other databases used to detect purely eye was created by me and consists of 349 photos of eyes, drawn from Internet (using the filter licenses has Google not to violate any copyright) and photos taken by the camera my mobile 13 Mpx. In the final part of the project the results are compared and conclusions are extracted, with which again the topic of possible applications and social impact they may have.