Loading...

Event Driven Camera Based Eye Tracking

Tajrobehkar, Mitra | 2014

663 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: English
  3. Document No: 46443 (55)
  4. University: Sharif University of Technology, International Campus, Kish Island
  5. Department: Science and Engineering
  6. Advisor(s): Jahed, Mehran
  7. Abstract:
  8. Robust detection of human eye has attracted interest during the past decade. Certain research studies have focused on the capability of determining the location of the human eyes while few have been focused on its tracking. The main task in this study is not only a system with eye detection capability but also a robust and non-intrusive eye tracker system of the eye movement using the iris as a trace element within the image captured through the camera. Such utilization is particularly applicable in diagnosing of certain abnormalities such as Autism which result in distinct abnormal movement and behavior in the eyes. Proposed eye tracker algorithm was implemented using the Kalman filter. The result was then utilized to solve the problem of finding the locations of gaze direction. The method of analysis was based on machine learning methods and video processing. Specifically, initially the face is detected within the image background using common preprocessing steps such as color space conversion, thresholding and Adaboost (Adaptive boosting) learning algorithm. Next, search areas for left and right eyes are determined using the geometrical properties of the face, histogram and SVM (Support Vector Machine) classifier. After the general detection of the eyes, through edge detection and Circular Hough Transform, the iris was identified and as a result, utilizing the geometrical location of the center of iris, the gaze information was extracted. Furthermore through tracking of the iris using Kalman filter, its coordinates and movement were evaluated. In summary, the study realized an automatic detection and tracking of eyes and irises through a RGB video sequence taken in a normal indoor environment without using any special lightings or complicated models. Based on obtained results, face, eye and iris tracking were performed with an average accuracy of 100, 99.4, and 98.4 percent, respectively
  9. Keywords:
  10. Hough Transform ; Kalman Filters ; Iris Movement Tracking ; Gaze Tracking

 Digital Object List

 Bookmark

No TOC