Multi-target Detection And Tracking Of The Panoramic Monitoring System

With the development of society,surveillance system is in a more and more important position in modern society.The monitoring system plays an significant role in road traffic,bank transaction,meeting site,meteorological observation,military reconnaissance and so on.A single function monitoring system has been unable to meet the needs of the development of modern society,how to achieve the high coverage and multi-functionality of surveillance system is the trends in future research and development.In view of this,after analyzing the achievements of previous research,this paper studies the method of panoramic generation,Multi-objective detection and tracking algorithm and video stabilization algorithm.Through the researches of imaging principles and imaging models of fish-eye camera,the imaging model of this paper's fish-eye camera is established;In view of the problem that the existing fish-eye image effective region extraction method's accuracy is not high and its running speed is not enough fast,An approximation algorithm based on median scan line is proposed;Then the analysis of existing fish-eye image distortion correction algorithm and their correction effects is finished;Aimed at the problem of poor correction effect on the existing fish eye image distortion algorithm,The distortion correction algorithm for fish eye image based on ellipse segmentation and distortion correction algorithm for fish eye image based on cylindrical expansion are presented.The principles of the conmon multi-objective detection and tracking algorithm in dynamic and static background is studied and their effects are shown,the their advantages and disadvantages are summarized;For the reason that the robustness of existing detection and tracking of multiple moving targets is poor and the tracking speed is slow,the multi-target detection and tracking algorithm based on average background and color histogram selective matching and the multi-target detection and tracking algorithm based on optical flow clustering and color histogram selective matching are put forward;The three indexes of target tracking rate,target retention rate and target error tracking rate are compared and analyzed between algorithms this paper proposed and the existing algorithms.The common methods of video image stabilization are analyzed,and the idea of electronic image stabilization is established;Due to the problem of poor stabilization and poor real-time performance of the existing electronic image stabilization algorithm,a video stabilization algorithm based on block matching and piecewise Calman filter is proposed.;Then the stabilization effect of the image stabilization algorithm based on the reference block and the segmented Calman filter is analyzed by the experiments.A panoramic monitoring system interface based on MFC is set up;The logical structure of three algorithms of distortion correction of fish eye image,multi-target detection and tracking and video image stabilization are optimized;The method of mapping table is used to accelerate the distortion correction speed of fish eye image;Then the sea state motion is simulated,and the monitoring effect of the panoramic monitoring system is verified;Finally,the GPU multi-thread programming method is used to ensure the real-time performance of the panoramic monitoring system.