基于Lucas-Kanada光流法的人眼特征点实时跟踪方法

Real-time Tracking Method for Eye Feature Points Based on Lucas-Kanada Optical Flow Algorithm

为更好地实现运动目标的检测跟踪,给出一种基于Viola-Jones特征点检测、亚像素角点提取与LucasKanade光流法跟踪相结合的人眼特征点实时跟踪方法。利用Viola-Jones算法检测出人眼区域,使用Harris角点算法在目标区域中提取亚像素级人眼特征点并引入筛选机制,从而在保证跟踪精度的同时减少运算量,采用金字塔分层机制的Lucas-Kanada光流法对运动的人眼特征点位置进行估计。该机制具有可切换的搜索窗口特点,兼顾了对大尺度高速度运动目标的跟踪。实验结果表明,该方法在保证算法实时性的同时明显提高了人眼特征点跟踪精度,并且能够保证快速及大尺度运动时的系统鲁棒性。

In order to realize the detection and tracking of moving targets,real-time eye feature points tracking method which combines Viola-Jones detecting algorithm, sub-pixel corner extracting with Lucas-Kanada optical flow algorithm is proposed in this paper. Viola-Jones algorithm is used to detect the eyes＇ rough region, and Harris operator with filtering mechanism is adopted to extract the accurate sub-pixel eye feature points in the target area,which can reduce the locating time but remain the locating accuracy. The Lucas-Kanada optical flow algorithm based on pyramid layer mechanism is introduced to capture the movement feature point. Taking into account the tracking for large scale and high speed objects, a switchable location windows method is presented and used in the last step. Experimental results show that this method can realize the eye feature points location and tracking for fast movement and large scale target with high precision, good real-time performance and strong robustness.