基于眨眼修正卡尔曼滤波的人眼跟踪研究

眼睛运动容易受到头部姿势变化、外界仿真干扰、实际光照条件等影响，已有眼部跟踪算法的准确率、鲁棒性较低。为此，提出一种基于眨眼修正卡尔曼滤波的人眼跟踪算法。采用垂直积分投影函数和水平积分投影函数得到人脸图像的眼睛区域，运用眼睛区域的颜色熵消除不相关因素，定位出瞳孔的位置，用卡尔曼滤波进行实时眼部跟踪，结合眨眼检测实时修正跟踪结果。实验结果表明，该算法准确率较高，实时性较好。

Research on Eye Tracking Based on Blinking Correction Kalman Filter

Eye movement is easily influenced by head posture change, external simulation interference and actual light conditions. The existing algorithms have low accuracy and robustness. An eye tracking algorithm based on blinking correction Kalman filter is proposed. Eye region of face image is detected using vertical and horizontal integral projection function. It uses the color entropy of eye region to eliminate uncorrelated factors, precisely locates the position of pupil. The eye region is tracked in real time by Kalman filter, and eye blink detection is combined to revise eye tracking. Experimental results show that the algorithm has high accuracy and good real-time.