Moving Target Detection And Tracking In Intelligent Video Monitoring System

The intelligent video monitoring system is an emerging field of the research of computervision, and in recent years,more and more intelligent monitoring system applied to all aspectsof people’s lives. The intelligent monitoring system using the technology of image processing,pattern recognition, machine vision and other related, to detecte collected moving target, tracktarget, as well as to understand and describe the target behavior.This topic sets up a monitoring platform which based on Haikang network monitoringcamera, by calling the improve SDK of Haikang to decode the video streams that transmittedover the network to achieve a real-time monitoring of the surveillance area, capture, andmonitor the areas of the saved video, and enables monitoring platform intelligence using targetdetection and tracking technology.The key technology of monitoring system intelligent is moving target detection andtracking technology. The purpose of moving target detection is extracting accurately the videopart we were interested in for the follow-up target tracking, identification and classification;tracking moving targets not only be able to accurately obtain the movement direction of movingtarget, speed, and other movement characteristics, also provide a reliable moving target data formoving target identification, classification, behavior, understanding and scene understanding.Information obtained by tracking the moving target, can also provide effective helps as thefeedback of the moving target detection and updating background model.The part of moving target detection,in this paper we analysis three moving target detectionalgorithm that are the optical flow method, the frame difference and background difference,and pointed out their advantages and disadvantages. This article focuses on a single Gaussianbackground subtraction and Gaussian mixture background model and the model updating. AndPut forward a new frame difference algorithm that is four inter-frame difference algorithm.Four inter-frame difference algorithm is relatively simple-to-use to achieve the ability to adaptto the environment, while the algorithm can effectively reduce the impact of the motion pictureof the history of the current moving target detection. Four frame difference algorithm also hasthe same defects of the traditional frame difference algorithm which are the moving target extracted is empty and the moving targets extracted is not full.As moving target tracking, in this paper we first provides commonly used moving targettracking algorithm.We highlights CamShift algorithm which is evolved by the MeanShifttracking algorithm based on target feature tracking and MeanShift algorithm, and these twoalgorithms are based on the color characteristics of the moving target of target trackingalgorithms, especially CamShift algorithm without parameters, and a small amount ofcomputing, do well in meeting a moving target for real-time tracking requirements.Semi-automated defect tracking algorithm for the CamShift be improved. A combination ofmoving target detection, automatic moving object tracking algorithm to achieve a good checkand track the moving target automatically without human involvement, intelligent monitoringsystem.