Robust Moving Object Detection For Intelligent Visual Surveillance

In Intelligent video surveillance, moving object detection is aimed at detecting if there is a moving object in video sequences from camera. The higher level application such as moving object classification,tracking,behavior understanding heavily depends on the results of moving object detection . However, moving detection is a hard work because of the change of background, such as influenced by the weather,illumination,shadow or outside interference. This paper is set in intelligent video surveillance, direct a true and complexity surveillance scenes, and discussing the mechanisms of outside interference which include illumination, shadow, small camera displacements and small motions such as tree branches motion. In order to enhance the robustness and stability of moving detecting, algorithms that include a real-time adaptive non-parametric thresholding algorithm,shadow removal algorithm,and suppression of false detection algorithm are put forward.Principles of moving detecting commonly used are discussed in detail. Examine and compare their merits and shortcomings. Based on the adaptive background mixture model, a real-time adaptive non-parametric thresholding algorithm for change detection is proposed in this paper. Based on the estimation of the scatter of regions of change in a difference image, a threshold of each image block is computed discriminatively, and then the global threshold is obtained by averaging all the thresholds for image blocks. The block threshold is calculated differently for regions of change and background. Experimental results show the proposed thresholding algorithm performs well for change detection with high efficiency,correctness and integral no matter indoors or outdoors.There are false detections due to random noise caused by the mechanisms of small camera displacements and small motions such as tree branches motion. This algorithm can adapt to illumination changes and suppress many non-static objects such as tree branch motion and similar small motion in scenes. Shadow detection is a outstanding and challenging question in moving object detection We analyze the property of shadow and the difference between the current image and background image of background regions,moving object regions and shadow regions. This paper presents an improved shadow detection method based on chromaticity information,luminance information and edge geometry information in YUV color space. Finally, a logic OR operation was performed on the above candidate shadow regions in order to remove the final shadow regions.Our experiments results prove that the proposed algorithm can effectively weaken the influence brought by the illumination changes or background confusion (typically such as the swag of the leaves) etc. and suppress shadow in moving object detection robust and timely.