Research Of Multi-Moving Object Detection And Tracking Based On DSP

Intelligent visual surveillance is one of the hotspots in the field of application for computer vision,which can detect, track and locate the moving object in the scene by the automated analysis of camera recorded video sequences without manual intervention. Moving targets detection and tracking are basis for intelligent visual surveillance.Our objective is to develop an algorithm for real-time multi-moving object detection and tracking and implement it on a DSP platform ultimately.The main work is as follows:Firstly, in part of moving detection, a great number of methods are introduced in the dissertation, including optical flow, temporal differernce and background subtraction method. The experiments show the performance of these methods. A improved background subtraction algorithm is presented in the dissertation, which integrates temporal difference with background subtraction method.The object's moving region is detected by background subtraction algorithm, the background model can be updated selectively.The case of background sharp change can be solved by temporal differernce method. A morphology filter is used to deal with the binarization image.The object can be segmented by the integral projection method.Secondly, in part of object tracking, by studying several related algorithm, a tracking method using a prediction which based on KALMAN filter is presented .It takes the target's histogram as the matching mask.We use the KALMAN filter to predict the position of object in the next frame, so the time for searching can be greatly reduced. The searching and matching for moving object are performed only in the predicted region. A approach of belief degree is used to deal with the case of combination and abruption when multi-moving objects are in the scene. The test result shows that the tracking algorithm can track multi-moving objects effectively and stably.Finally, we describe the DSP platform—DM6437 IP camera, and discuss the tips about algorithm porting and optimization.