瞳孔检测的图像裁剪与异常瞳孔排除

为获取眼部特征进行视线跟踪，提出了一种具有较高实时性和正确率的瞳孔检测方法.使用模板匹配定位眼睛在图像中位置，裁剪边缘冗余图像，降低瞳孔检测计算量，从而提高实时性；利用瞳孔的形状、颜色等特征以及眼球运动规律，得到瞳孔在图像中分布规律，对错误的瞳孔信息进行排除，从而提高瞳孔检测正确率.实验结果表明，在NVIDIA Jetson TX2嵌入式计算机上，该瞳孔检测方法检测正确率达到95.06%，检测速率为95 fps，耗时平均减少55.33%，具有良好的实用性. 

Image Cropping and Abnormal Pupil Exclusion for Pupil Detection

In order to obtain eye features for line-of-sight tracking, a pupil detection method with high real-time performance and accuracy was proposed. Real-time performance was improved by cropping the redundant edge image with template matching to locate the position of the eye in the image, and reducing the amount of pupil detection calculations. The shape and color of the pupil and the rules of eye movement were used to obtain the distribution rule of the pupil in the image. The distribution rule was excluded to improve the accuracy of pupil detection. The experimental results show that on the NVIDIA Jetson TX2 embedded computer, the detection accuracy of the pupil detection method reaches 95.06%, the detection rate is 95 fps, and the time-consuming average reduction is 55.33%, which has good practicality.