Research On Algorithms Of Moving Object Detection And Tracking In Intelligent Video Surveillance

Intelligent video surveillance technology has been widely applied in life, business, national security and military applications and other areas. The range of the research on intelligent video surveillance is very wide, including moving object detection, moving object tracking and other parts. This thesis focuses on the research on the algorithms of moving object detection and tracking.In the process of moving object detection, traditional background modeling algorithm is prone to get the problems of slow update rate of background model and difficult matching of background model. Aiming at these problems, this thesis proposes an improved algorithm based on background modeling of Gaussian Mixture Modeling. Based on the original model, the algorithm introduces a background model update parameter, and by comparing the value of the parameter and current pixel value to determine whether the current pixel is background pixel or not. Meanwhile, the algorithm also sets a lower limit for standard deviation parameters of Gaussian Mixture Model. It can not only improve the number of matching models, but also can reduce unnecessary new constructed models. Experiments show that the accuracy of object detection of the improved background modeling algorithm has improved to some extent.In the process of object tracking, due to changes in illumination, traditional Mean Shift algorithm is prone to get the problem of deviation of tracking rectangle. Aiming at this problem, this thesis proposes an improved algorithm based on MS algorithm. The algorithm introduces SURF which is invariant to illumination change. By SURF feature extraction and matching in both object region and candidate region, then calculate the offset of the candidate region. The algorithm also introduces the Bhattacharyya coefficient to compare the accuracy of results from MS algorithm and SURF tracking method, and chooses the maximum degree of accuracy as the final result. Experiments show that in the case of illumination changes, the improved algorithm can track moving objects more stably.Due to changes in the size of moving object, it’s prone to get the problem of deviation of tracking rectangle during object tracking. Aiming at this problem, this thesis introduces affine transformation model to the improved MS. By calculating affine transformation parameters of each frame, the improved MS with affine transformation determines the size and position of the moving object, and then adaptively adjust the size of the tracking rectangle. The algorithm introduces the Bhattacharyya coefficient, and compare the above results with the others obtained from the improved MS and SURF method, then chooses the maximum degree of accuracy as the final tracking result. Experiments show that in the cases of illumination change and object’s size change, the improved MS with affine transformation can track object more stably.