Recognition And Tracking For Flight Target Based On DSmT

Flight target recognition and tracking as one of the important research contents in the field of machine vision and pattern recognition, has important significance in identification, defense intelligence, automatic gun and other military areas. In the process of flight target recognition and tracking, because of the high speed, distance, complex background and other factors, flight target recognition and tracking plays a prominent role and is a hard and challenging task. This thesis takes flight target recognition and tracking as the study's core and has a detailed in-depth study to detection, recognition and tracking.As we all know, the general enhancement algorithms have the problem of missing target detail information. In view of this, this paper presents an algorithm which is stratification adaptive enhancement based on wavelet. The algorithm calculate the appropriate contrast information based on the low-frequency part of wavelet decomposition, to achieve adaptive slicing, then get the enhancement function base on adaptive hierarchical results, finally leads to keeping detail information while enhancing. In order to meet the needs of real-time and segmentation, a new algorithm based on one dimensional maximum entropy thresholding and corner detection is used. According to the characteristics of the image, the algorithm can automatically choose the segmentation algorithm, which can ensure the real-time performance of the segmentation algorithm.In view of the problem of low accuracy of target recognition using single feature, combined with Dezert-Smarandache theory(DSm T), this paper develop an algorithm to identify the target class through the fusion of three sources of evidence. The algorithm fusion classification information of the support vector machine on the target histogram of oriented gradient and pairwise geometric histogram, and construct the corresponding evidence source with the information about the target velocity. The Simulation results show that the fusion recognition method is more accurate than single target recognition method.Finally, in order to solve the problem of multiple target tracking is easy to loss when the target track is close to or even intersect, combined with DSm T, this paper uses an algorithm based on multi information fusion. The algorithm takes into account the target's motion information and the target's properties. Firstly, the position of the target is predicted by using the kalman filter. Then the observation value of the position is analyzed. Finally, the accurate target location is determined. Using this algorithm can solve the problem of multiple target tracking is easy to loss when the target track is close to or even intersect.