基于DSP的快速目标识别与跟踪技术研究仪器科学与技术专业论文.docxVIP

哈尔滨工业大学工学硕士学位论文 哈尔滨工业大学工学硕士学位论文 - - II - Abstract With development of intelligent systems and analysis techniques, various image recognition systems have gained widely attention both in military and civil field. Especially, moving object detect has been an important research subject because of its wide application prospect. The subject “Study on fast moving object detect and tracking based on DSP” is focus on the development of fast and accurately detect and tracking moving object based on DSP technique. Firstly, the paper developed two kinds of detect and tracking approach separately for dim object and general large object. Due to the long distance, the target may appear as a dim point and has no feature information like shapes and textures. Furthermore, aero-optic disturbances and air turbulence make the SNR of a single image very low in the real environments. All of above make the miss distance of the missile difficult to estimate in shooting range. In order to solve these problems, a novel method for detecting and tracking small moving point target in image sequence is proposed in this paper, based on combination of the correlation of multi-frames and wave gate techniques. Especially, the qualifications to confirm the target and the establishment of the wave gate based on size and characteristic moving parameters of the target are stressed in the paper. Meanwhile, objective to the requirement of both real-time and accuracy for moving shaped target detect and tracking, a novel method is proposed based on combination of images difference and feature invariant. The paper improved the tradition difference method, constituted a feature invariant which characterized the object contour without influence of object’s size change, shifting, and rotation, and developed a dynamic-extremum-matching scheme. Secondly, the implementation of detect algorism based on DSP is researched, including the driver design of input and output, the realization of JPEG code and network commu