| In recent years, information technology has a rapid development. Thus, its application has used in every field of people life. People take advantage of communication of these relevant advanced technology to service for our human all aspects like our society, economic, resource, environment and so on, and in the activity, human-computer interaction is always the theme. In view of the fact that the importance of the technology, research on how to realize a natural, convenient and ubiquitous human computer interaction has become the highest goal of modern information technology, its innovation will lead the new trend of technology. Thereby, the eye tracking algorithm has very significant application in the field of human-computer interaction, and there also exists a great space for development. Besides, in the research field of medical, the accurate location of pupil and some eye moving tracking method also can contribute to certain eye disease, and to some mental disorders of the brain, it can also provide an effective treatment, make a convenient new and diagnosis method for doctor. And even to the whole medical community, it also a novel creation. Otherwise, in the aspect of authentication, the biological identification system always developed quickly. The complication of the network world, make our personal information is easily stolen and leaked. So, the identification of various occasions are getting higher and higher. Recent years, the research of iris recognition system has attracted more and more attentions, and more and more researchers begin to realize that the accurate location is always a part cannot be ignored.As far as the current public algorithm, it can be seen that a great improvement and breakthrough has occurred in this field, but it also exists a lot drawbacks. There is still much room to improvement.Like the existence of image processing noise, such as off-axis aberrations, physiological distortion and occlusion of hair and eyelids, often lead to suboptimal algorithms. However, although there has some research in these aspect, but it cannot achieved an ideal result. Therefor the pupil location algorithm also faced a great challenge. In addition, the processing of pupil image is always important, to ensure a high performance of the algorithm, it also need ensure a high quality of the image.In the light of the insufficient of current public algorithm, the paper proposed a high precision, high efficiency and robustness pupil location algorithm based on combination of physiological characteristics of the human eye and advanced technology in the field of information technology. At first, we used the device introduced at second chapter to extract the eye moving video, then a frame of iris image is read from the HD eye motion video file, and compressed it to a pixel of 768*342. Meanwhile we apply the Adaboost classifier that trained by Haar features of the image to detect and extract the region of interest(ROI). Next, to detect and fill the reflections of the extracted region of interest. Then the algorithm used the starburst model to detect the feature points of pupil edge, applied clustering method based on high density connected region to cluster these points. Finally, random sample consensus(RANSAC) algorithm is used to fit the ellipse fitting of the pupil edge. At experiment part, in order to evaluate the performance of the algorithm, we have made a statistics on the accuracy and the average execution time of each frame image, and compared the results with the current popular algorithms. Through the comparison, we can know that our algorithm has a high performance. |
PDF Full Text Request