Based On Gaussian Mixture Background Modeling And Shadow Suppression Algorithm

In computer vision and intelligent video surveillance community, the higher levels such as moving object classification, tracking and behavior understanding heavily depend on the results of moving object detection. Moving object detection is one of the fundamental tasks for video analysis, which is active topic widely researched in the last decade around the world. In this thesis, we focus on the background modeling of complex environment and moving object detection, and place emphasis on mixture Gaussian model(GMM) and shadow suppression especially. The main contributions are as follows:1. To our knowledge that mixture Gaussian model theory is not analyzed systematically and the variant algorithms are not compared each other comprehensively, through analyzing numerous of reference papers, we research GMM through systematical theoretical derivation and experimental analysis.2. Several background modeling algorithms are widely researched and analyzed based on extensive experiments, and the methods are compared each other in speed, memory requirement and accuracy. Finally, the comparison conclusions are presented.3. The main methods of shadow detection and suppression are researched systematically, furthermore, they are classified and compared each other through experimental analysis.4. An improved mixture Gaussian background modeling algorithm is presented to reduce the computational cost of traditional mixture Gaussian algorithms. The improved mixture Gaussian-based background model updates the parameters of Gaussians according to the frequency of a pixel value changes. Experimental results show that our algorithm can improve the processing speed greatly and detect moving object accurately.5. A new mixture Gaussian-based clustering algorithm is proposed to suppress moving shadow. This method firstly decides whether a pixel value belongs to probable-shadow, which is then put into mixture Gaussian shadow model to learn and cluster. The experimental results indicate that the proposed approach can run in real-time and remove shadow effectively.