Pupil Detection in Real-World Environment
| Autor: | WU, JIA-YAN, 吳佳燕 |
|---|---|
| Rok vydání: | 2018 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 106 Understanding the movement of the eyeball has always been the focus of various research fields including medicine, psychology, advertising, and application control. In the past, there have been many studies on eye tracking in the interior. However, due to the evolution of the times, there are more and more portable eye tracking applications using outdoor scenes, such as sports, shopping, and driving. There are many ways to achieve eye tracking, and the most widely used one is the pupil tracking using a head-mounted infrared camera.There are many challenges in pupil tracking in real-world scenes, such as lighting changes, eyeglass reflections, make-up, etc. The use of a head-mounted eye tracker can change the angle of the shot due to the shape of the person, the height of the nose, and the height of the ear. In this paper, edge detection and edge segmentation techniques are used to separate the edge segments and use statistical and pupil conditions to find the pupil center. If the edge detection cannot detect the pupil, further use the grayscale value to binarize and find the pupil. This paper optimizes the accuracy when the pupil is obscured by the eyelids. This article uses four credible databases, each with three different cameras, for a total of 226,705 labeled pupil images for verification. In the experimental results, the accuracy rate reached 74.6%. When the resolution is 388x288 and 640x480 respectively, the FPS is 36 and 25. We believe that this speed can handle most cameras. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|