摘要
基于图像分割以及原目标检测的视觉跟踪是一种极具潜力的跟踪方法,该方法首先对视频中的图像序列用多种分割方法进行图像分割;然后基于概率潜在语义分析(PLSA)算法对分割区域进行类别估计;再从这些区域中筛选出各自的候选区域,对其进行选择性结合,提取出原目标,从而获得原目标的空间信息.最后基于原目标的空间信息,用贝叶斯算法对目标进行跟踪,并采用EM算法来优化跟踪算法.实验证明,该方法优于其他的跟踪方法,能鲁棒地处理遮挡,分散和光照变化等问题.
Visual tracking based on image segmentation and proto - object detection is a potential tracking ap- proach. Firstly, it gives an image sequence based on multiple segmentations, and uses PLSA to classify each segment region ; then it obtains the final proto - objects based on selecting the candidate regions from these segment regions for a selective combination. Finally, the target is tracked with the Bayesian approach based on the spatial information of the protoobjects from the previously obtained proto -objects. It uses the EM algorithm to optimize the tracking method. Experimental results demonstrate that the proposed approach achieves good performance which can robustly deal with occlusion, distraction as well as illumination variation.
出处
《云南民族大学学报（自然科学版）》
CAS
2015年第6期496-500,513,共6页
Journal of Yunnan Minzu University:Natural Sciences Edition
基金
南京航空航天大学研究生创新基地(实验室)开放基金(KFJJ201426)
