Study Of Maneuvering Target Tracking Method Based On Binocular Vision

Maneuvering target tracking is an important research topic in the field of computer vision technology and it has a wide range of applications in many areas, such as robot navigation, aerospace, manufacturing process control, medical research, intelligent surveillance and military reconnaissance. The so-called mobile target tracking is that first to detect and identify the moving targets from a sequence of images that contains the moving tagets, and calculate the target displacement, velocity and acceleration and other information, then use the appropriate algorithm to forecast and track the moving targets. However, the relatively mature mobile target tracking technology are mostly under the single vision model, so it limits the application of maneuvering target tracking technology in three-dimensional scene. So, this paper studied the moving target tracking technology which based on computer binocular.vision system.In this paper, with the use of disparity in stereo vision, the object was accurately detected and extracted by the difference of real-time disparity and background disparity, so the affection of shadow was well eliminated and the object was accurately detected and extracted. The method in this paper was proved to be able to accurately extract moving object, comparing with the result of traditional background difference. For the moving targets prediction and tracking, this paper focused on the interactive multiple model algorithm and CamShift (Continuous Adaptive Meanshift) tracking algorithm which based on the color histogram. First, a deep research on the interacting multiple model (IMM) algorithm-one of the comprehensive adaptive filter was proposed. The simulation results showed that the IMM algorithm which contains maneuvering target "current" statistical model is superior to the basic IMM algorithm in maneuvering target tracking. Then the CamShift tracking algorithm based on color histogram is described in detail, and its strengths and weaknesses were analyzed, on this basis a new maneuvering target tracking algorithm combined the IMM and CamShift was proposed, named CamShiftIMM algorithm. Experiments showed that the new algorithm achieved the fast and accurate tracking, and solved the target occlusion to some extent. Finally, the experiments for the moving targets tracking of the general movement, random mobility, and short-time occluded movement were designed on the binocular vision system, using the CamShift algorithm and CamShiftIMM algorithm. Experimental results proved that this new method used in the thesis is not only able to track the general moving object and the maneuvering mobile targets quickly and efficiently; but also able to track the slow-maneuvering target and the maneuvering target of unknown motion law .In summary, the new algorithm is robust and practical.