Research On Key Image Processing Algorithm Of Rotational Optoelectronics-guiding Missile

With the development of the military technologies, now, the short-distance interceptorequipped point-type infrared seeker cannot accommodate the needs of the modern war. Thesmall rotary missiles which have advantage in simple and reliable system, compactstructure, cost-effective and good anti-interference ability are applicable especially to lowaltitude and super-low altitude short-distance interceptors. The electro-optical imagingguided seeker is essential to increasing the work range and the adaptability to the complexwar background, and it’s a development tendency of the small missile for terminalinterceptors. So, the dissertation of this research has important significance foraccommodating the needs.There are three difficulties in the development of the electro-optical imaging guidedseeker of the small rotary missiles: the first is the small physical dimension limited by thewarhead, the second is the image stabilization, and the third is the objects detection andtracking. This dissertation concentrates its research on the image processing algorithm forthe electro-optical imaging guided seeker of the small rotary missiles. The researchconcerns three key technology fields: image stabilization algorithm for large motion offset,moving objects detection and objects tracking. The goal of this dissertation is the practicalapplication of the algorithms and technology. Therefore, a solid foundation for thedevelopment of the electro-optical imaging guided seeker of the small rotary missiles canbe laid. Four innovative contents researched in the paper are as follows:Firstly, when a camera is mounted on the high speed motion platform, the videosequence that it outputs change rapidly. A motion estimation method which combining thepolar coordinates transform in the frequency domain with the Gauss-Newton minimizationin space domain is presented to accurately stabilize the video sequence with large motionoffset.. At first, a rough offset value between two frames’ images is coarsely estimatedusing the polar coordinates transform in the frequency domain, Then the accurate offsetbetween two frames’ images is efficiently estimated using Gauss-Newton minimizationmethod by taking the rough offset as the initial value. Therefore, a stable video sequence by image compensation using the accurate offset is achieved. Experimental results show thatthe optimized algorithm can accurately stabilize the video sequence with large motionoffset in real time.Secondly, a new approach to detect multiple objects in the image sequences usingmultiple particle filters is presented. Filter over the sequence of difference images in spaceand time fields using the particle filters based Bayes rules to eliminate the noise and clutter,and detect the moving objects using the particle clustering algorithm according to theproperty that the particles will converge in the region which the objects locate at. Anapproach based on Kullback-Leibler distance(KLD) is introduced for improving theefficiency and reducing the computational complexity of particle filter. The approach todetect multiple moving objects using multiple adaptive particle filters is validated by theexperiments.Thirdly, to solve to the problem that the traditional Mean-Shift algorithm has poordescription of the feature space to the object area, an improved Mean-Shift object trackingalgorithm is proposed. Utilize a description of the feature space combining the image grayfeatures and enhanced edge features, which can enhance the tracking robustness when theobjects and background’s color is similar or the image has low contrast and great noise.According to the problem that the traditional Mean-Shift algorithm does not have atemplate update mechanism, the LMS (Least Mean Square) filter for updating adaptivelythe object kernel histogram is introduced to solve the long time stable tracking problem.According to the problem that the traditional Mean-Shift algorithm cannot adjustadaptively the kernel bandwidth, the Mean-Shift object tracking algorithm fixes the object’sscale adopting improved successive bandwidth trials to solve the object size change.According to the problem that the traditional Mean-Shift algorithm cannot converge to theright location when the object moves rapidly because of the absence of a predictionmechanism for the object location, we integrates the improved Mean-Shift tracking methodand Kalman filtering to enhance the tracking ability for the fast moving objects. Resultsfrom a lot of experimental images verify the effectiveness of the improved algorithm.Finally, all algorithm introduced by this dissertation has been refined and integratedinto the electro-optical imaging guided seeker of the rotary missiles, and been real-time processed to the image of the seeker. Results from a lot of experimental images verify theeffectiveness of the algorithm introduced by this dissertation for image stabilization, objectsdetection and tracking.