结合目标估计的自适应压缩跟踪

Adaptive Compression Tracking Algorithm Combined With Target Estimation

传统的压缩跟踪算法使用不考虑目标位移的搜素机制和固定大小的跟踪窗口跟踪目标,且每帧更新分类器,容易造成目标的漂移和丢失.该文提出了结合目标估计的自适应压缩跟踪算法,该算法采用ORB算法将先前帧和当前帧的目标区域进行特征点匹配,通过建立关系模型实现候选目标的估计定位,从而解决了搜索区域受目标运动速度影响的问题;其次,提出了基于匹配点间方差的尺度自适应方法,实现了跟踪窗口的自适应变化;最后,引入遮挡判断因子,并结合匹配点的数量调整分类器的更新策略,实现抗遮挡的处理.实验结果表明:该算法具有较高的重叠率和成功率,对目标跟踪中的相机晃动、快速运动、尺度变化、遮挡等因素具有较强的抗干扰能力,且在多个测试视频上平均帧率为32帧/s,能满足实时性的要求,较好的解决了传统压缩跟踪算法的缺陷.

The traditional compression tracking algorithm uses the search mechanism without consideration the target displacement and the fixed size tracking windowto track the target,and each frame updates the classifier,which is easy to cause the drift and loss of the target. This paper propose an adaptive compression tracking algorithm combined with target estimation. the ORB algorithm is used to match feature points of the target area between the previous frame and the current frame,By the relation model to estimate positional candidate target,so as to solve the influence of the search area by moving target speed. Secondly,this paper proposes a scale adaptation method based on the variance between matching points,which realize the adaptive change of the tracking window;Finally,introducing occlusion judgement factor,combine with the number of matching points to update strategy of adjusting classifier,which realize anti occlusion processing. The experimental results showthat the algorithm has high overlap rate and success rate,and has strong anti-interference ability for target tracking in camera shake,fast motion,occlusion,scale changes and other factors. Moreover,the average frame rate is 32 frames/s in a number of test video,which can meet the requirements of real-time. and it can solve the shortcomings of the traditional compression tracking algorithm.