研究了一种改进的基于Hough变换圆检测的瞳孔识别方法。直接利用Hough变换圆检测进行瞳孔识别准确度低、运算量大, 为此, 本文提出了改进的识别方法。在实现Canny边缘检测的基础上, 利用Hough变换进行求解时, 限定检测的半径范围, 在此范围内, 根据参数空间圆的方程, 求出对应的圆心坐标, 得到最佳拟合圆, 最终实现瞳孔识别以及中心点定位。在vc++6.0开发平台下, 对150幅图像进行瞳孔识别检测实验, 结果显示, 直接利用Hough变换检测方法的准确率为70.7%, 平均速度为1 s; 本方法的准确率为94%, 平均速度为0.45 s, 可见相较于直接利用Hough变换进行瞳孔识别, 本文方法在识别准确率和识别速度方面均有显著提高。结果表明, 此方法能够快速而准确地进行瞳孔识别以及中心点定位。

As traditional way for pupil recognition based on Hough transform circle detection has lower accuracy and speed, an improved method is proposed in this paper. After edge detection by Canny operator, Hough transform is performed in a limited radius. Within this range, finding the best fitting circle based on the circle equation in parameter space to get the center coordinates and radius, pupil recognition and center positioning are achieved ultimately. 150 images are used to test the effects of pupil recognition under vc ++ 6.0 development platform. The experiments results show that the accuracy is 70.7% with an average speed of 1 s based on Hough transform which is used directly. In comparison the accuracy is 94% and the average speed is 0.45 s with the method proposed in this paper. It indicates that this method has a significant improvement on accuracy and speed. It can recognize the pupil and locate the center faster and has more accuracy.