提出一种基于直方图二值化和改进的轮廓跟踪的瞳孔检测方法。首先对采集的瞳孔图像计算直方图, 根据直方图特征取出合适的阈值进行二值化, 再对二值化后的图像进行轮廓跟踪, 得到精确的轮廓边界, 进而得到直径、中心等信息。对前一帧瞳孔图像进行瞳孔信息计算后, 后一帧图像以该瞳孔中心为中心, 缩小检测区域进行直方图二值化及瞳孔轮廓跟踪, 得到该帧图像的瞳孔信息, 减少运算, 加快检测速率。该算法鲁棒性高, 速度快, 可满足 300帧/s的实时检测速度。

This paper presents a new pupil detection method based on histogram binarization and improved contour tracking. Firstly, the pupil image histogram is calculated to get the suitable threshold for binarization. Then contour track is used to get the precise contour. Finally the information such as diameter and the center of the pupil is attained. According to the pupil information of the previous frame, the detection region for histogram binarization is narrowed and improved contour tracking is used to get the pupil information of the next frames. These methods will reduce the amount of computation and speed up the detection rate. The algorithm is robust, fast, and can meet the real-time detection speed of 300 frames per second.