The Research For Recognition And Tracking Of Moving Vehicle Based On Visual

Visual signal have the advantage of wide detection range and rich information accession. With the rapid development of image processing technology and computer processing ability in these years, the visual navigation has broad application prospect in intelligence vehicle's navigation. With the supporting project of National Scientific and Technical Fund in Shanxi, the research on moving vehicle recognition and tracking method based on the visual is put forward in this thesis, which provides profound research to image recognition and tracking method.In traditional vehicle recognition, the contour recognition of the vehicle has been achieved by the tracking method based on the visual. Traditional detection method of frame-difference method realize the detection purpose through comparing of continues two frame in sequence image, but this algorithm make inaccurate detection also by detecting the background area (revealing background) revealed in current frame (last frame is covered by target). For the problems above, improved second finite difference algorithm based on frame-difference method, which rely on the difference between three neighboring images and could determine the location of moving vehicle in the middle image based on the method of"and"and"binary"two difference results, is proposed in this thesis considering the computing efficiency and accuracy. This algorithm has high computing efficiency and eliminates last revealing background covered by target vehicle, which not only continue the advantages of frame-difference method, but also overcome its disadvantages.In traditional vehicle tracking, when the vehicle is turning or part of it is blocked, traditional matching model methods easily get the error-matching and tracking lost problems, and have low computing efficiency. In terms of problems above, a kind of Face tracking method, CamShift algorithm is used in this new field. CamShift algorithm is a kind of tracking methods based on target color characteristics, which does not affected by the shape,size changing of target, and efficiently solves the error-tracking problems when vehicle turn or part of it is blocked.On the basis of the vehicle recognition and tracking, the visual ranging model is build combined with the optical and geometry principle, relying on the tracking results of CamShift algorithm, which was testified that it could realize vehicle ranging.The algorithm has realized system's function based on the OpenCV programming in the Visual C++ integrated development environment. The recognition and tracking of the vehicle on the road and radio control car is achieved. Experiment results show that second finite difference method could realize vehicle location recognition accurately. CamShift algorithm can track moving vehicle accurately when the vehicle turn or part of it is blocked, which effectiveness is confirmed in the moving vehicle recognition and tracking based on the visual.