Research On Synthetic Aperture Radar Scene Matching

As a branch of image registration, synthetic aperture radar (SAR) scene matching is the process of registering two images of the same place and determining the space relation of them. Existing scene matching techniques broadly fall into two categories: the area-based and feature-based methods. In this paper, we present two contour based scene matching algorithms: a closed-contour matching method based on hausdorff distance and an open-contour matching method based on chain code correlation. We also give a hierarchical area-based matching method.In the area-based method, usually a small window of points in the first image is statistically compared with windows of the same size in the second image. The measure of match is usually the normalized cross-correlation. However area-based methods like this are time consuming, and a hierarchical area-based matching method is presented in this paper so that the computing time is reduced exponentially. This method works well for L-band and C-band SAR image pairs taken at different time.In the closed-contour based method, this paper presents a closed-contour extraction and contour-matching scheme for optical-to-SAR scene matching problem. The given closed-contour extraction method is based on bottom-hat transformation, erosion and dilation techniques in mathematical morphology. Hausdorff distance is a method for measuring the resemblance of two point sets, which compute the maximum-minimum distance. Hausdorff distances between the long contours are computed, and those contour pairs of low-hausdorff-distance are assumed to be matching contour pairs. So we can get the image transformation parameters.An open-contour based method is also presented in this paper. First...