基于双边滤波的协方差目标跟踪算法  

Covariance object tracking algorithm based on bilateral filtering

作　　者：谢英红[1,2] 李凌[3] 何宇清[1] 

机构地区：[1]天津大学电子信息工程学院,天津300072 [2]沈阳大学信息工程学院,沈阳110044 [3]唐山师范学院,唐山063000

出　　处：《电子测量与仪器学报》2016年第8期1282-1287,共6页Journal of Electronic Measurement and Instrumentation

基　　金：国家自然科学基金(61503274);国家自然科学基金(61472274);辽宁省教育厅科学技术研究项目(L2014481);辽宁省博士启动基金(201501090)资助项目

摘　　要：现有的使用协方差建模的目标跟踪方法在目标形变或是光照变化较大的情况下,达不到理想的跟踪效果。在分析目前协方差建模目标跟踪方法缺点的基础上,提出一种融合双边滤波的协方差目标跟踪算法。首先,将待跟踪图像进行双边滤波,提取滤波后的图像特征构建协方差特征矩阵作为跟踪模板。其次,由于协方差矩阵为对称正定流形,服从对数-欧几里德黎曼度量。在此度量下,构建了目标协方差矩阵相似性度量和模板更新策略。各种条件下的实验结果表明,新构建的基于双边滤波的协方差特征矩阵对目标形变和光照变化有更好的适应性。The existing object tracking method using covariance modeling is hard to reach the desired tracking performance when the deformation of moving target and illumination changes are drastic. Based on the analysis of the shortcoming of target tracking method using covarianee modeling, a covariance object tracking algorithm based on bilateral filtering is proposed. Firstly, the algorithm deals the image to be tracked with bilateral filtering, and extracts the needed features of filtered image to construct covariance matrix as tracking model. Secondly, covarianee matrix is positive symmetric manifold, so it obeys log- Euclidean Riemannian metric. Under this metric, the similarity measure for object covariance matrix and the model updating strategy are constructed. Extensive experiments show that the proposed covariance matrix based on bilateral filtering has better adaptability for object deformation and illumination changes.

关 键 词：目标跟踪 双边滤波 协方差矩阵 对数-欧几里德 黎曼度量 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TN141.9[自动化与计算机技术—计算机科学与技术]