A Study On The Technology Of The Infrared Sequences Image Matching Tracking

The infrared image matching tracking is the one of the key technology in the domain of the infrared imaging homing. In the project practices, as a result of the pose change of the detector flat, the variety of the distance between the missile and the target, the change of the target pose, the disturb of the background wave and the influence of the imaging noise, the grey and geometry distortion complicatedly are brought in the image target, and the extremely trouble of the image matching tracking are produced. Such being the case, the research of the infrared image matching tracking have numerous challenges.In the paper, for the problem of the infrared matching tracking in the image, two kinds technology of the matching tracking was studied, which were the the matching technology between image frames and the the matching technology possessed of the translation, rotation and scaling.In the paper, the mostly work and the innovational achievement were:(1) In order to solve the image matching errors of the image target rotate, a mean removal correlation tracking algorithm in the arbitrary specified image target of complicated ground background was presented. In the algorithm, after the computing formula of the correlation coefficient was deduced approximately, the disadvantage of the great operations of the correlation coefficient has been solved. And in the algorithm, the adaptive template image update method and the modificatory adaptive template image update method were presented., and the performanences of the two methods was compared.(2) In order to solve the problems of the image noise and the local shelter in the process of the image matching, the adaptive image matching tracking algorithm based on the corresponding pixel distance measurement was presented. In the algorithm, the matching tracking confidence factor was designed, and the template image was adaptively updated with weighting of the matching tracking confidence factor.(3) The image information theory and the property of information entropy was studied. The concept of the mutual information, the information entropy and the cross-entropy was indroduced into the image matching. The four kinds of the information measures are that the maximum mutual information, the minimum join entropy, the normal mutual information and the entropy correlation coefficient. was studied. The theory analysis shows: when the image matching of the noise distributing and the grey transform is ambiguity, on the condition of the former template update, these information measures has possessed good matching performance. On the basic, the image matching tracking algorithm based on the unit-entropy vector feature and the local cross-entropy were presented.(4) The image matching tracking algorithm based on the gradient methods(GM) was studied. On the embody: first, the GM, SGM, BDGM algorithm was introduced; and the iterative solved methods of these algorithm's movement parameter estimate was introduced, then the convergence property of the three algorithms was studied. Secondly, the two methods of constructing the matrix H was presented, and the former gradient methods was implemented, the contrast experiment was achieved, the better matching effect was acquired.(5) When the less templat and the biggish range of translation, rotation, scaling and noise were found between the template image and observation image, in order to implement the matching between images, the image matching algorithms based on the Fourier-Mellin transformation and based on the Newton iterative were presented. Through experiment, the convergence property and adaptability of the Newton iterative algorithm was analyzed.