Moving Object Detection And Shadow Removal Based On Video Analysis

Intelligent video surveillance system is an important application in computer vision, moving object detection is the key technology in intelligent video monitoring system and shadow removal is an important step in moving object detection.Combining the thoughts of image intensity normalization and two-dimensional cross-entropy, a moving object detection algorithm, which is robust against illumination changes, is presented in this paper. Firstly, the intensity of each video frame image is normalized. Secondly, the background is initialized by an improved mean filtering method, which can get a satisfying background image although an object appears. Then the background is updated automatically. Finally, the moving objects are detected by using an adaptive threshold based on two-dimensional cross-entropy to segment the background subtraction image. By combining invariant color features and edge detection technique, a new shadow removal algorithm is proposed in this paper. First, the shadow detection result based on invariant color features is verified by Gaussian Mixture Model. Secondly, the edge image of the object is obtained combining edge detection technology. Finally, the object contour is extracted using a shrinkable active contour model to get the moving object without shadows. Experimental results demonstrate that the proposed algorithm is effective in moving object detection.