The eye tracking system plays an important role in clinical medicine,in order to raise the recognition rate of the eye characteristic value in the system.A double filtration algorithm is innovatively proposed to locate the center of pupil and the center of corneal reflection in this paper.First,the acquired eyes image was processed by the pupil center location algorithm to reach binarization,and then the contour points with 100 ～300 was set as the threshold.Next,pupil contour was ellipse fitted using the least square ellipse fitting method,and the contour that was most close to the circle was used as the pupil contour to obtain pupil center.On this basis,the edge of the eye image in the certain range of the pupil center was detected,and then the external rectangular center of the contour was calculated.Taking the center of the pupil as the starting point for coordinate transformation,the corneal reflection spot center after converting coordinates was eventually calculated.This algorithm not only reduced the area of image processing to a certain extent,but also improved the processing speed and the recognition rate of reflection spot.The experimental results showed that the designed algorithm accurately located the pupil center and corneal reflex center in different conditions in real time.The maximun nean square deviation of pupil-corneal reflection vector was less than O.74 pixels,and the average error was less than 0.53 pixels.The algorithm run at a speed of 12.8 ms/frame.The algorithm met the precision as well as showed very strong robustness,being of significance for the development of eye tracker.%眼动追踪系统在临床医学中有着重要应用,为提高系统中眼图特征值的识别率,创新性地提出一种双层过滤的瞳孔中心和角膜反射光斑中心的定位算法.瞳孔中心定位算法首先对获取的眼睛图像二值化处理,然后以100 ～300为阈值设定轮廓点数,进行第一次瞳孔过滤,最后用最小二乘法椭圆拟合瞳孔轮廓,并将最接近圆的轮廓作为瞳孔轮廓,继而得到瞳孔中心.在此基础上,进一步对瞳孔中心一定范围内的眼图进行边缘检测,求其轮廓,并找出轮廓的外接矩形中心,然后以瞳孔中心为基点进行坐标转换,最终求出角膜反射光斑中心.算法在一定程度上缩小图像处理的区域,提高实时性和反射光斑的识别率.实验结果表明,所设计的算法能够实时地对不同情况下的瞳孔和角膜反射光斑中心进行准确定位,瞳孔-角膜反射向量坐标最大均方差低于0.74像素,平均误差低于0.53像素.算法运行速度达到12.8 ms/帧.算法满足精确性的同时具有很强的鲁棒性,为后续眼动设备的开发提供重要的参考价值.
展开▼
