Research On Multi-Camera Target Tracking System And Platform Development

Target tracking is one of advanced research topics in computer vision,which uses the image processing, pattern recognition and computer vision technology, extracts and analysises the key information of video sequence, detects and deals with the abnormal cases in monitoring scenario timely.It plays a very important role in multi-agent cooperation, continuous tracking of suspicious target, visual navigation robot buffet, intelligent transportation monitoring&traffic violation tracking and driverless in military field especially in intelligent monitoring.With the development of hardware and software technology, intelligent monitoring steps to network and intelligent direction from the intitial single camera.More and more researchers have joined in the study of multi-camera intelligent monitoring.It is the basis of intelligent monitoring to realize the multi-camera target tracking, and how to implement the object handover in multi-camera system is the key problem.The automatic generation of scence MAP in multi-camera tracking is a hard problem.This thesis mostly studies the tracking target handoff problem and does exploratory experiments on the generation of scence MAP,and at last develops a platform that can be used to make multi-camera target tracking experiment based on the network.Firstly, this thesis makes a conclusion of the main research methods and technology abroad in recent years, and uses the memebership concept of fuzzy logic to quantify the color feature of tracking target. Then the color blob model is set up combining with the prior geometric proportion relation and the motion features and the model is used to the target initialization automatically.The results based on the model indicate that this method can locate the target automatically when the target color and texture feature is simple. The article also makes research in the basic link of multi-camera target tracking, the target tracking in the single camera, and realized the target detecting and tracking intelligently in the single camera scence.In the terms of multi-camera target tracking and handoff,combining with the geometric topology of scence model this thesis makes a deep research in the generation algorithm of scence MAP automatically and a large number of experiments. The algorithm can realize the antomic learning of camera topology and the generatiin of scence MAP by determining at the moment when the target in and out the scence and the probability.The preset function of PTZ camera is used for quick location of holder.The experiment result shows that this method has realized fsat accurate judgement and schedule of the handoff camera, and accomplished the task of relay.This thesis developed a multi-camera target trackin system based on the hardware and software of PTZ cameras, and all the system is connected through LAN. Target tracking is completed by the video sever which is connected to the camera.All the control informations are transited by the management server and the video data can be transfered directly in the net..In the thesis, the encapsulation of general module avoids the waste of iterative development. At the last of4th chapter, the experiment results show that the system is successful and it complete the task planed.