The Research On Eye-tracking Based On Template-matching And Lucas-kanade

The computer vision has been experienced a rapid development science the 70s of 20 century, especially in the 21st century, with the continuous improvement of computer performance and low-cost digital image capture technology. In the field of computer vision, eye detection and tracking has been attracting more and more researchers'interest. As one of the important feature in facial expression, the tracking on the eye can help computer to recognize and understand the human's expression, and this research will promote the development of the affective computing in turn. In this paper, we handle the eye images of the video sequence, and then use the dynamic object tracking and recognition algorithm to realize algorithm we proposed. In order to improve the defaults of the existing algorithms, in this paper a new methods been bring out based on the template-matching and Lucas-Kanade algorithm. The main thesis works are as follows:(1) The face detection and estimating eye region based on the geometry. Firstly we study the Haar-like features and Adaboost learning algorithm; based on these algorithms we get the head image. Then based on the priori knowledge, we extract the eye region.(2) Recognizing and tracking the eye. In the traditional methods of the moving object tracking, template matching method is the most intuitive and widely used, but there is a compute-intensive, low efficiency problem. While the Lucas-Kanade algorithm with the image-based pyramid algorithm can solve this problem. So in this paper we combine the advantages of both, use of template matching to find the eye of the initial image of the sequence, then tracking the eye by the Lucas-Kanade algorithm. (3) Eye state identification. For each eye region we get from the frame, binary the eye region by the Otsu method, get the eye's vertically integrated projection by the gray integral projection method, and finally get the eye's station by the projection curve. (4) From the above study of eye tracking, we realize the automatic identification system, and we also analyzed the experimental results. Summarize the advantages and disadvantages of this method, laid a goof foundation for the future research.