Moving Object Detection via Robust Low-Rank and Sparse Separating with High-Order Structural Constraint

Low-rank representation has been successfully applied for moving object detection by assuming the background images are linearly correlated while the moving foreground are sparse. Further, extensive works propose to incorporate the spatial pairwise smoothness of pixels to improve the robustness. In this paper, we investigate the long-range spatiotemporal relationships among pixels, and propose a novel approach to pursue the high-order consistency for moving object detection in the low-rank and sparse separation framework. In particular, we integrate the sparse unary penalty, the spatial pairwise smoothness, and the supervoxel-based high-order consistency into a unified structural constraints on the foreground. Moreover, we propose a single optimization algorithm to learn the background model and the foreground mask at a same time. Extensive experiments on the benchmark datasets GTFD and CDnet suggest that our approach achieves superior performance over several state-of-the-art algorithms.