SAR Image And Optical Image Of Registration Research

Image registration is a different time, different perspectives or different sensors in the same scene two or more images matching and fusion process, in other words,determine transformation between the matching image and the input image, makes them on the best matching in the space.This paper studies SAR image and optical image registration. SAR image is a synthetic aperture radar image, optical image is collected by the traditional image equipment, because of the different image principles, both of them contain the different information, the registration of them can give full play to their advantages, overcome their limitations.Because the SAR image mechanism is special, feature extraction is very complicated, different noise and different resolution of SAR image have different texture structure, conventional image matching algorithm is very difficult to obtain a better matching effect,the heterologous registration research between SAR image and optical image is necessary, the main work of this paper:(1) analysed the basic principle of image matching, raise mathematical description and induction of the image matching, advance the four elements composition of matching algorithm; SAR image preprocessing, because SAR image have speckle noise, filter is needed, comparing several filtering algorithm,select the best performing filtering algorithm,prepare for the realization of algorithm.(2) studied matching algorithm based on SIFT algorithm, SIFT algorithm is described and extracted local feature point of image,it is not relevant to the image scaling and rotation angle and has very strong robustness.The feature of space description has greatly improved,a better description of image feature extraction has a better effect, in order to reduce the calculated amount, this paper studies SURF algorithm based on matching,its feature points are less than feature points of SIFT algorithm, it improves the speed of image matching,robustness and high matching precision, The algorithm is more efficient, faster and better in real-time character than SIFT algorithm. (3) extract feature points in use of Harris corner detection algorithm introduce the Mahalanobis distance of similarity criterion, achieve the image matching, implementation of image matching,the experiments show that the algorithm is effective. introduct all kinds of improved Hausdorff distances, and take the LTS-Hausdorff distance as a measure of image matching similarity criterion, analysed the basic principle of genetic algorithm, summary components of genetic algorithm, proposed one kind based on the Hausdorff distance and genetic algorithm for image matching algorithm.The algorithm take the image edge as the feature space,, firstly, dichotomize the decomposed images, secondly genetic algorithm as the search strategy, is transformed into solving the global optimal solution.Experimental results show that the algorithm is effective.