基于自适应相关滤波器的人眼定位改进方法

人眼定位是人脸识别、视觉跟踪系统中一个关键步骤。针对传统人眼定位方法定位精度低、误差稳定性差的问题，提出将自适应相关滤波器与灰度积分投影法相结合，在训练和测试阶段做了两点改进。首先对训练得到的自适应相关滤波器，在\-0.2,-0.1,0,0.1,0.2\]角度范围内旋转得到5个滤波器，选择灰度最大值对应的位置作为初始定位点，然后在该定位点的5×5邻域内，水平和垂直方向分别做灰度积分投影，最后选择积分最小值的位置作为最终的目标定位点。与平均合成精确滤波器、最小输出平方误差和滤波器、自适应合成相关滤波器算法相比，改进算法的定位精度最大提高2.9%，平均绝对误差和标准差均低于原始算法。实验结果表明，提出的改进算法在定位精度和稳定性上均优于以上算法。

An improved eye location method based on adaptive correlation filter

Eye location is one of the most important steps in face recognition and visual tracking system.For the problems of lower eye location accuracy and unstable error distribution based on traditional eye location methods,the adaptive correlation filter is combined with integral projection to detect eye position precisely,and two improvements are made in the training and test phases.Firstly,adaptive synthetic correlation filter(ASCF) is rotated within the angle range of -0.2,-0.1,0,0.1,0.2,and the corresponding location of the maximum grey value will serve as an initial anchor point,and then gray-level integral projection is performed in the horizontal and vertical direction respectively within the anchor point 5×5 neighborhood.Finally,the integral minimum value is used as the ultimate goal orientation.Compared with the average synthetic exact filter,minimum output sum of squared error filter and adaptive synthetic correlation filter,the position accuracy of improved algorithm is increased by 2.9% and the mean absolute error and standard deviation are lower than those of original algorithms.Experimental results show the proposed improvements are superior to the general algorithms in terms of accuracy and stability.