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

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.