Research On Multiple Moving Object Detecting And Tracking In Comolex Scene

With intelligent surveillance playing an increasingly important role in people’s life, video sequences motion analysis has become an important research topic. Moving target detection and tracking is one of the key technologies which catches widespread attention and has become a hot topic recently.In this paper, the research object is the video sequence which is captured by a static camera and it researches on the issues about moving object detection in complex background. First, it introduces the background and present situation of the object detection and tracking study, and leads to the focus of the study, including object extraction, shadow elimination, target tracking and other methods. And the feasibility of the algorithm is verified by experiments and comparison.In the object extraction, this paper lists several typical methods used in object detection and determines the background subtraction as the method for moving objects extraction. Then it introduces several common background modeling methods, focusing on a background extraction method based on the intrinsic image. By repeated experiments and adjustments we has achieved the background model. Finally by combining differential image with the one on gradient field, thus a moving target is detected effectively. In the shadow detection, this paper summarizes the various shadow elimination methods. According to the shadow edge features and LBP texture feature, combining with morphological processing, the shadow is better eliminated.In the target tracking, this paper compares several universal application of the moving target tracking methods. Based on the moving object detected in previous paper, we use Kalman filter to predict the target location in the next frame and get the forecast range. Finally, we use the matching algorithm to match and achieve a better result of Multi-target tracking.