Total Variation Regularized OR-PCA for Foreground Detection in Water Monitoring
Author:Chunxue Wu,Hejie Chen, XiaoLi,YanWu, Naixue Xiong Author Unit: School of Optical-Electrical and Computer Engineering, University of Shanghai for Science and Technology, Shanghai, China. Public and Environmental Affairs, Indiana University Bloomington, IN 47405, USA. Department of Mathematics and Computer Science, Northeastrn State University, Tahlequah, OK, USA.
Abstract:In the water monitoring system, we need to detect some abnormal conditions for security protection in real
time. For example, the suspicious person walks back and forth, and the equipment of water pump house runs
abnormally. On this issue, the existing Robust Principal Component Analysis (RPCA) methods is difficult
to deal with some complex dynamic background. Furthermore, with the increase of the number of video
frames, the calculation and storage capacity of RPCA method is also increasing, which is not conducive to
process big data. Therefore, three methods are proposed to solve the above problem in this paper. Firstly,
the spatial continuity of the foreground is used to separate the foreground from the dynamic background
and the temporal continuity of the foreground is used to detect irregular objects movement. Secondly,the
frame difference method is used to remove redundant frames in video. Thirdly, this paper presents a
foreground detection method based online robust principal component analysis (OR-PCA). To deal with
this decomposition, a minimization problem, stochastic optimization is employed. Experimental results
demonstrates that the proposed method can detect foreground accurately in complex background.
Keywords:OR-PCA; The frame difference, Spatial continuity; Temporal continuity; Stochastic optimization
Download information [SIZE:1.23 MB Amount of downloads: second]