Moving object detection by robust PCA solved via a linearized symmetric alternating direction method

Abstract

Robust Principal Components Analysis (RPCA) gives a suitable framework to separate moving objects from the background. The background sequence is then modeled by a low rank subspace that can gradually change over time, while the moving objects constitute the correlated sparse outliers. RPCA problem can be exactly solved via convex optimization that minimizes a combination of the nuclear norm and the l 1-norm. This convex optimization is commonly solved by an Alternating Direction Method (ADM) that is not applicable in real application, because it is computationally expensive and needs a huge size of memory. In this paper, we propose to use a Linearized Symmetric Alternating Direction Method (LSADM) to achieve RPCA for moving object detection. LSADM in its fast version requires less computational time than ADM. Experimental results on the Wallflower and I2R datasets show the robustness of the proposed approach. © 2012 Springer-Verlag.