The Research Of Moving Object Tracking And Retrieval In Surveillance Video

With the improvement of safety awareness, intelligent monitoring system has beenmore widely used in traffic management, security management, and public places. Inwhich a large number of surveillance videos urgently need an intelligent, efficient systemto deal with. In recent years, moving target extraction, tracking and retrieval have beenpart and parcel of researches on surveillance video.It analyze the advantages and disadvantages of some common methods used toextract moving targets in surveillance video. Then using bounding box method deal withthe no-connected object. Taking advantage of the color channels which are the mostdifference between the target and the background, Mean-shift algorithm would be morerobust. In the combination of Kalman filter and Mean-shift algorithm, Kalman filter couldpredict the state of every tracker. The tracker in good state will track only using Kalmanfilter. When occlusion occurs between the targets, Mean-shift will track every target in theocclusion area. Algorithm show its robustness and efficient.At the end, this paperintroduces the concept of color space and color difference metrics, and then describes amethod for extracting a representative color subset from the color set of moving targets. Incolor retrieval, the color subset will be presented to user, so an exact value is not needed.The color in subset could show color of moving targets in general.