Moving Object Detection And Tracking Based On Video Images Sequence

Image processing is one of the most popular research directions in the computer science.It is widely and importantly applicable to traffic video monitoring, aerospace, robot vision,medical image analysis. So this subject is considered as one of the most developmentprospective subjects at present. Although a lot of research has already done by some scholars,there are a number of question unresolved yet. So there is important theoretical significanceand practical value to research this subject. In this paper, common target detection andtracking algorithms are further improved based on the base of image sequence motion targetdetection algorithm and target tracking algorithm.In this thesis, the domestic and foreign researches development of moving targetdetection and moving target tracking based on image sequence are introduced first. At thesame time, some relevant preprocessing used in image processing such as grey leveltransformation, binaryzation, Image denoising and mathematical morphological processing.This provides theory basis for the following improvement of algorithm.In the moving target detection, lighting flow, background subtraction, inter-framedifference are detailed introduced in this paper. Their advantages and disadvantages arecompared, as we know lighting flow requires the support of hardware condition, thetraditional second frame differential method is easy to cause the shortcoming of movementtarget area being drafted, out of shape and several large areas empties. Three framesdistraction may cause to not getting the clear edge of the moving objects.Consideing thesequestions, five frames distraction is put forward to get more precise and clear edge of movingobjects, the problem of large area empties is improved in large degree.In moving object tracking, at present, mean shift is usually found in the field of objecttracking. Its shortage is that it will lose efficacy when the object is sheltered. In this thesisKalman filter, Extend Kalman filter, Mean shift algorithm, Shadowing filter algorithm areintroduced. It proves by simulate experiment that the function of shadowing filter is superiorto EKF, so it is superior to kalman filter. To solve the problem that the effect and precise oftracking moving target of sheltering with Mean shift algorithm was unsatisfactory, the traditional algorithm was improved by predicting the target location in next frame of videowith Kalman filter or shadowing filter. It shows by experiment that the algorithms improvedhave high reliability and robustness.