一种基于MCD的抗遮挡快速跟踪算法

A Rapid Algorithm of Tracking Objects under Occlusions Based on MCD

针对遮挡、目标形变、噪声干扰等复杂场景下目标跟踪问题,提出了一种基于最大邻近距离(Maxium Close Distance,MCD)的抗遮挡快速跟踪算法,定义了灰度差异和最大邻近距离相结合的相似性度量方法,并根据相关系数自适应更新目标模板,利用多十字交叉快速搜索方法加速目标匹配的过程。当目标被遮挡时,采用Kalman预测滤波算法预测目标位置,维持目标的跟踪。实验结果表明,该算法能实现遮挡情况下的目标跟踪,目标跟踪精度优于8个像素,为遮挡、目标形变、噪声干扰等复杂场景下目标实时跟踪提供了一种解决方法。

To solve the problem of tracking objects under occlusions and noise,a rapid algorithm of tracking objects under occlusions based on MCD（Maximum Close Distance）is proposed in this paper.A similarity measurement is defined based on grayscale differences and MCD.Adaptive template update strategy which can adjust the correlation coefficient is used to improve the stability of object tracking.Cross search strategy is used to speed target up matching.Kalman filter is applied to predict the object＇s trajectory under occlusions.Tracking experiments indicate that this algorithm succeeds to track objects under occlusions and the target tracking accuracy is better than 6 pixels.It provides a solution to the problem of real-time tracking of targets in complex scenes such as occlusion,target deformation and noise jamming.