Study On The Key Technologies Of Moving Object Detection And Tracking

The key technology of moving target detection and tracking is studied in thisthesis, major problems of moving target detection and tracking algorithm facing isanalyzed and discussed in-depth in conscientiously sum up the results of previousstudies. Thesis study two key technologies, one is moving image segmentation, thesecond is moving target tracking based on particle filter.In order to obtain the initial background image and update it from image sequencewhich contains moving object, a lot of literature at home and abroad is in-depth studied.mean method, median method, statistical histogram, sparse Bayesian classifiermethods is used to obtain the initial background; Kalman filtering method, Gaussianmixture models is used to update the background. Aimed at the deficiency of largecalculation and difficult shadow removed of traditional Gaussian mixture model, Thispassage proposes the Gaussian mixture model combined with block idea based onHSV space, which is reduced the computational cost and eliminated the shadow of smoving object. the performance of the proposed method of moving objectsegmentation is verified.In recent years, robotics, target tracking, and many other applications pay moreattention on particle filter. On the base of proposal distribution and adaptive choicemechanism, this passage proposed two methods based on particle filter. One is basedon hybrid proposal distribution and adaptive resampling of particle filter to improvemoving object tracking algorithm; the anther algorithm is improved adaptive proposaldistribution of proposal distribution in diversify. The simulation program platform usesmatlab7.0, and the result is feasible.