适用于遮挡问题的目标跟踪算法

Object Tracking Algorithm Applied to Occlusion Issue

提出一种基于网格模型的目标跟踪算法 该算法首先进行遮挡区域检测 ,然后进行网格结点的运动估计和网格更新过程完成目标的多帧跟踪 改进的遮挡区域检测算法有效地提高了检测准确度 ,从而确保遮挡区域的准确跟踪 ;网格结点的运动估计是通过特征窗口运动补偿匹配完成 ,可以有效地克服块效应 实验证明 ,该算法解决了二维运动估计时网格模型在遮挡区域存在的问题 。

A novel object tracking algorithm based on mesh model is proposed to perform accurate object tracking while occlusion problem exists. Occlusion detection is done first, and then mesh nodes motion estimation and mesh update are carried to achieve object tracking in several frames. An improved occlusion detection method is applied to enhance the detection precision. Mesh nodes motion estimation is performed through feature window motion compensation matching, which can overcome block artifacts. Experimental results indicate that the proposed algorithm can solve occlusion problems existed in 2D motion estimation, and can accurate track object effectively.