Moving Object Detection via Robust Principal Component Analysis

The technology of intelligent video surveillance that can replace human operators to monitor the areas under surveillance is becoming more important in many fields of modern society. This paper proposes a novel approach for background and foreground segmentation and moving objects detection which can be used for video surveillance. Unlike general approaches, this algorithm constructs observation data D and uses robust principal component analysis which inherits the idea of traditional Principal component analysis and is the newest development of matrix sparseness and compressive sensing to compute low-rank matrix A and sparse matrix E. By matrixizing the each column of matrix A and E, background and foreground segmentation can be finished. The process of moving objects detection based on foreground matrix E is also present in this paper according to the histogram specialty of sparse matrix. The experimental results demonstrate that the algorithm can finish background and foreground segmentation, and moving objects detection efficiently from image sequences.