一种新的鲁棒主成分分析方法及其应用  

A novel robust principal component analysis method and its applications

作　　者：陈甲英 赵建伟[1] 曹飞龙[1] 

机构地区：[1]中国计量学院理学院,浙江杭州310018

出　　处：《中国计量学院学报》2016年第1期113-120,共8页Journal of China Jiliang University

基　　金：国家自然科学基金资助项目(No.61571410;61272023)

摘　　要：背景建模在视频运动分析中具有重要作用.视频序列背景图像通常具有低秩性,为了更好地刻画该特性,精确提取视频背景,提出了一种基于截断核范数的鲁棒主成分分析模型.同时设计了一种两步迭代算法来求解该模型,最后将该算法应用于视频背景建模.不同视频数据库实验表明,该算法对于求解背景建模问题是有效的.Background modeling is critical in the video motion analysis.To describe the low rank property of a set of video sequences and to improve the accuracy of background modeling,a novel robust principal component analysis with the truncated nuclear norm method was proposed.A model with a two step iterative strategy was used to solve the problem.Finally,the algorithm was applied in the video background modeling.A series of experiments on different video databases demonstrated that the algorithm was effective for the solving of the background modeling problem.

关 键 词：矩阵恢复 截断核范数 鲁棒主成分分析 背景建模 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]