多特征稳健主成分分析的视频运动目标分割

提出一种多特征稳健主成分分析(MFRPCA)算法,该算法融合多种视觉特征进行视频运动目标分割,分割的目的即将运动目标从静止信息中提取出来,分割的主要过程是将多特征视频矩阵分解为低秩矩阵和稀疏矩阵.矩阵分解过程是求解一个带受限条件的核范数与L2,1范数组合的最小化问题,此最小化问题可以通过增广拉格朗日乘子法(ALM)有效求解.与其他算法相比,本文算法融合了图像的颜色、边缘和纹理特征等多个特征,通过对变化检测基准数据集进行检测,本文算法获得的查全率为0.486 0和F度量为0.559 7,实验结果表明,本文算法的稳健性和可靠性均优于其他算法.

Multi-feature robust principal component analysis for video moving object segmentation

In this paper, we propose an algorithm named Multi-feature Robust Principal Component Analysis(MFRPCA), to integrate multiple visual features for Video Moving Object Segmentation. The aim of Video Moving Object Segmentation is to separate the moving objects from the static information. Its main process is to decompose the multiple feature video matrices into low rank and sparse matrices. The decomposition is to solve a minimization problem formulated as a constrained combination of nuclear norm and L_2,1 norm, which can be solved efficiently by Augmented Lagrange Multiplier(ALM) method. Compared to other methods developed recently, the proposed method integrates color, edge and texture features. The quantitative results for Recall and F-measure obtained from experiments on the change detection benchmark dataset are 0.486 0 and 0.559 7 respectively. The results which outperform other methods well show that the proposed MFRPCA can achieve more robust and reliable performance.