中文摘要：

针对多目标检测和跟踪过程中常发生的合并和分裂现象,提出一种对不同运动目标假设不同的运动模型,并基于隐马尔科夫度量场（HMMF）的检测和跟踪算法.为了更准确地估计目标的运动参数,还提出了一种简单有效的单目相机标定算法.仿真结果表明,文中算法对遮挡不敏感,即使在发生遮挡、合并或分裂情况时也能很好地跟踪运动目标.

英文摘要：

This paper presented a detecting and tracking muhiple motion objects method. The method designs dissimilar motion models for different objects which often merge and split each other in the surveillance video. The method based on hidden Markov measure field（HMMF） is robust for lost information. A simple and efficient calibration of monocular camera was also presented to get more accurate motion parameters. The simulation results show that the algorithm can successfully track objects even if the occlusion happens.