| Synthetic Aperture Radar (SAR) is significant in many fields such as geological exploration and surveillance, etc, and many researchers pay more attention to this area of relative technologies about SAR target matching recognition. In this paper, some methods on speckle suppression, target segmentation and match are studied.Basing on multiplicative model of speckle, Frost, Kuan, Lee and Gamma MAP filters are researched, they are all on the hypotheses that the mean of speckle is stationary and estimate real image in stated neighborhood, so the size of the window can make big influence on the effect of speckle suppression. The methods based on relative standard deviation and based on edge preservation are studied. The experiment indicates that the former can accomplish speckle suppression effectively, and the latter makes good effects on holding edge information of SAR image.On the basis of spatial relationships of the targets, the model images are classified as four kinds of array, harbor, airdrome and blob target. The method of Ostu and area threshold expurgation are researched. The improved method of double parameters CFAR is also studied which is on the hypotheses that obeying Rayleigh distribution and makes the mean and variance of the ring like window as parameters and completes target segmentation. The conventional Hough Transform for linear feature extraction of airport target is improved, and a method of area Hough Transform is presented. The experiment indicates that Ostu is a simple, fast algorithm and makes good effect to array targets, the method of area threshold expurgation can remove small spot area aroused from speckle or background clutter, the improved CFAR gets good segmentation to blob target fast and the area Hough transform method accomplishes line extraction in images which contains multiple targets.The area of harbor target is available after segmentation, seven moment invariants are get and the match between model image of harbor and measured image is accomplished by normalized Euclidean distance; After line extraction from harbor and airdrome target, the geometrical features of the lines and gray- scales of the support areas are acquired and the match between model image and measured image is accomplished, then the conclusion is that the line match method based on support area and the method based on normalized Euclidean distance both possess upper success ratio and lower false detecting rate by the experiment results, but the former is better than the latter in the integrated effect. |
