Moving Object Detection Based On Low Rank And Sparse Decomposition

Moving object segmentation is one of the basic problems of visual analysis, has the widespread application prospect in the event detection, video retrieval etc. This proposal aims to construct a new structured low rank representation model for video data, which is then applied to the video motion object segmentation problem. The model decomposes the background and foreground as the low rank and sparse part of the video matrix respectively. The low rank prior of background and the temporal-spatial continuity prior of foreground motion objects is fully used to segment the motion objects robustly.The robust principal component analysis model fails to handle the long time and large scale video data. A moving object detection method based on hierarchical RPCA is proposed in this paper. The first-pass RPCA rapidly identifies the likely regions of foreground in the down-sampled video sequence. Motion salience map is then generated based on the motion saliency of these foreground regions to achieve rapid and efficient detection of moving targets.The robust principal component analysis model fails to effectively use the spatial-temporal continuity priors of moving object, thus tends to misclassify dynamic details in the background as moving target problem. This paper presents a method of moving target detection using spatial and temporal information. A3-D spatial-temporal total variation model filters the unstructured background disturbance in sparse component. Motion salience map is then generated based on the motion saliency of these foreground regions. Second-pass builds a weighted RPCA model, which impose a weighted threshold of candidate moving target. The weighted RPCA model makes the foreground detection robust and can obtain clear and complete foreground. Experimental results show that this method can effectively handle complex dynamic background of moving target detection.