基于最大似然准则Hausdorff距离的跟踪算法  

Probabilistic Hausdorff Distance Used for Tracking Objects

作　　者：何力[1] 傅忠谦[1] 郑重[1] 

机构地区：[1]中国科学技术大学电子科学与技术系,安徽合肥230027

出　　处：《计算机仿真》2008年第2期232-236,共5页Computer Simulation

摘　　要：针对视频处理中运动物体的检测和跟踪问题,提出一种基于最大似然准则Hausdorff距离的目标跟踪算法,首先利用基于GVF的Snake方法获得物体模型;然后采用基于最大似然准则的Hausdorff距离匹配后续帧中的目标,搜索策略采用类似于Rucklidge提出的多分辨率搜索方法,在不影响搜索成功率和目标定位精度的情况下,可以显著地缩短搜索时间;最后使用Snake方法完成运动物体的轮廓更新。实验表明该方法可以较好地跟踪刚性和非刚性物体,同时对部分被遮挡的目标也有良好的跟踪效果。This paper proposes a new method of tracking moving object based on maximum likelihood hausdorff distance. First the object model is detected using snake. For object tracking, the contour is used to estimate its motion in the next frame by maximum likelihood hausdorff distance, and for searching strategy, Rucklidge＇s methods are used, thus a much higher search speed can be achieved while maintaining the search success rate and target location accuracy. And then a modified snake is used to update the object model. Experiment results show that the proposed algorithm can track moving object efficiently, and it also can track partially occluded object accurately.

关 键 词：最大似然准则 豪斯多夫距离 运动目标跟踪 

分 类 号：TP371.4[自动化与计算机技术—计算机系统结构]