基于显著特征区域和概率图模型的目标跟踪

针对目标跟踪中的遮挡问题,提出一种基于局部显著特征区域和概率图模型的跟踪算法.提取目标的一组局部显著特征点,以局部显著特征点为依据,在目标区域中划分出多个感兴趣的显著特征区域;利用这组显著特征区域相互之间的空间位置关系,并结合各个显著特征区域自身的局部信息建立目标的马尔可夫随机场(MRF)模型;采用mean shift(MS)算法对各个显著特征区域分别进行跟踪,利用MRF模型对MS算法的跟踪结果进行概率推断,融合各个显著特征区域的权重,精确定位运动目标的最终位置.在多个视频序列上的实验结果表明,与改进的MS算法、粒子滤波算法以及分块跟踪方法相比,此算法具有较高的跟踪精度;尤其是当目标被遮挡时,该算法具有较好的跟踪鲁棒性.该算法充分利用了显著特征区域自身的局部特征和区域之间的空间结构信息,能够实现复杂情况下的运动目标的鲁棒跟踪.

Object tracking based on salient feature regions and the probabilistic graphical model

Occlusion is one of the most challenging problems in the object tracking community. To deal with the occlusion problem,this paper presents a local salient feature- based probabilistic graphical model for visual tracking.Firstly,salient points of the visual object are extracted,and then regions of interest( ROIs) of the object,derived from the salient points,can be obtained. Secondly,combining spatial constraints among different ROIs and the local information contained in each ROI,the object is represented as a probabilistic graphical model. Finally,based on the object model and the Mean Shift tracking results of each ROI,the probabilistic inference algorithm is adopted to estimate the probability of each ROI belonging to the given object region. Comprehensive experiments are carried out on several testing video shows,and compared with three well- known trackers,i. e. the improved Mean Shift,the Particle Filter and the Fragments- based tracker,the proposed method has a higher tracking accuracy and robustness,especially under the occlusion condition. The proposed tracker,using the information of local saliency and spatial structure for tracking the object effectively,can perform with high robustness in complex real-world scenarios.