Study Of Video Object Tracking Technology Based On Wireless Network

Along with the the increasing of society demand and development of video processing technology,A new generation of video surveillance system can realize the real-time transmission of high-definition video through the network,and is more and more widely used in reality;and the target tracking algorithm as the focus in the field of computer vision,has very important practical value and application prospect,the realization of intelligent monitoring system with the function of target tracking has become a hot issue.In this paper,an object tracking system based on wireless network is realized with the knowledge of video compression,transmission,image processing and so on.The design of the system is divided into client and server,the client is divided into Cloud Terrace client and mobile robot client,the client mainly realizes H.264 format encoding of the collected video,WIFI real-time video transmission to the server,the server mainly completes the data decoding,the processing of specified target by tracking algorithm,real time monitoring of multi-channel video,and send control command to client.Tracking algorithm is an improved kernel correlation filters algorithm with adaptive learning rate according to occlusion and model changes.This paper mainly completes the following work:1.The software design uses the C/S structure,the client in the windows and Linux system to complete the video acquisition,H.264 format video compression and the wireless transmission function,and can contro the movement according to server control command.2.The server is built on the Windows system of the PC,is MFC procedures developed by using C++ language in the VS2010 environment,using the multi thread way to complete the reception and decompression of multi-channel video,the real-time video display function,and can complete the processing by target tracking algorithm for target tracking,and indicate the Cloud Terrace client and mobile robot client according to the instructions of the server.3.Kernel correlation filters algorithm,STC algorithm and CT algorithm are compared through the multi group experiments,the improved kernel correlation filters algorithm will be applied to the system according to the experimental results and data analysis.4.In the improved adaptive update Kernel correlation filters algorithm based on the original Kernel correlation filters algorithm,the detected occlusion and changes of model is fed to a classifier model updating,all previous frames are also used in the classifier model updating.,and for multi camera and long time tracking the classifier model is corrected according to the relocation method.The experimental results show that the system designed in this paper is feasible and has good tracking performance.