基于随机森林的长期目标跟踪方法

Long-term Target Tracking Method Based on Random Forests

结合正负样本相互作用思想和随机森林算法构建检测器,融合基于LK光流法的跟踪器,提出一种基于TLD(Tracking Learning Detecting)的随机森林长期目标跟踪方法。将该方法与Mean-Shift算法、TLD算法进行对比,结果表明该算法能很好应对目标丢失、遮挡情况,准确率在93%以上。在多种情况下对该方法进行实验验证,可实现刚性物体和非刚性物体在复杂背景下的长时间精确跟踪。

In this paper, we propose a target tracking method based on Tracking Learning Detec-ting （ TLD） random fores by using the detector constructed by the ideas of the interaction be-tween positive and negative samples and random forest algorithm, and tracker based on LK opti-cal flow method. This method is performed the comparison with the Mean Shift algorithm and TLD method. The results show that the algorithm can have the strong robustness to target lost, target occlusion, and the accuracy rate is more than 93%. The experiment results in many cases verify that this method can achieve a long time accurate tracking for rigid and non-rigid object in complex background.