类属图密集近邻搜索的视觉跟踪算法研究

摘要

A visual tracking algorithm based on dense neighborhoods searching on affinity graph,which can be applied in the scenes with large deformations and severe occlusions.Object parts are extracted based on image segmentation model of Markov random field,affinity graph matrix is buih by searching dense neighborhoods in affinity graph matrix,confidence map of target location is obtained by matching relationship to determine location of target tracking.Experimental results show that the proposed algorithm is better compared to other state-of-the-art methods.%提出一种基于密集近邻搜索的视觉跟踪算法,能够有效应对目标跟踪过程中出现的形变和遮挡问题.基于马尔科夫随机场建立图像分割模型,提取出目标部件,建立目标部件的类属图矩阵;通过搜索类属图矩阵中的密集近邻,得到相邻帧之间目标部件的匹配关系;通过匹配关系得到跟踪目标位置概率图,确定目标跟踪位置.实验结果表明:本文提出的方法相比其他同类方法效果更好.