基于分布场的特征融合重检测鲁棒性跟踪算法研究

摘要

论文针对分布场目标跟踪算法中使用分布场的目标模型估计鲁棒性较弱的问题,提出了一种将分布场与其他特征有效融合的方法.在对每个像素点进行分布场估计时,原始算法仅通过该点的灰度直方图来估计其在灰度空间上的分布,并没有考虑该点的位置与结构信息.为了实现在分布场中对目标的结构信息的有效表示,本文通过对目标中包含结构信息的特殊点进行特殊编码实现特征融合.此外,引入失败检测机制提高算法的精确度.实验证明,在光照变化的情况下,融合了结构信息的分布常比原始分布场在目标跟踪的效果有很大提升,且优于当前流行的目标跟踪方法.%This paper proposes a method to integrate the distribution field with other features effectively,aiming at the prob-lem of weak robustness of the target model estimation of in the distribution field target tracking algorithm.When estimating the distri-bution field for each pixel,the original algorithm only uses the gray histogram of the point to estimate its distribution in the gray space,and it does not consider the location and structure information of the point.In order to realize the effective representation of the structure information of the target in the distribution field,in this paper,the feature fusion is realized by special coding of spe-cial points containing structural information in the target.In addition,the introduction of a failure detection mechanism improves the accuracy of the algorithm.Experiments show that in case of changes in illumination,the distribution of structural information fusion is often more effective than the original distribution field in target tracking,and is superior to the current popular target tracking method.