IMPLEMENTATION OF EYE DETECTION USING DUAL CAMERA ON THE EMBEDDED SYSTEM

摘要

This paper presents an implementation of eye detection on the embedded system. Two camera systems based on the low cost Raspberry Pi modules are employed. To speed up the process, the proposed system implements the face detection technique and the eye detection technique on two camera modules separately. The face detection module detects the bounding box of face and sends the coordinates to the eye detection module via a serial communication. In the eye detection module, the eye is searched on a limited area defined by the face's bounding box. The popular Viola-Jones object detection is employed in the face detection module. Three eye detection techniques, consisting of the Viola-Jones method, the eye-map method, and the Hough circle transform method, are implemented and evaluated in the eye detection module. The best result is obtained by the Hough circle transform method, where the frame rate of 30.020 fps, the true positive rate of 0.869, and the precision of 0.824 are achieved.