False-alarm reduction in mine classification using multiple looks from a synthetic aperture sonar

An algorithm is proposed for performing underwater mine classification when provided with multiple synthetic aperture sonar (SAS) views of each object. The proposed method consists of three main steps. First, the multiple images of a given object are registered to a common reference frame. Second, the registered images are fused to form a single image, thereby allowing all of the information contained in the multiple images to be exploited simultaneously. Third, the fused image is compared to a library of image-templates of targets of interest. The approach is demonstrated on one data set of simulated SAS images and on one data set of real, measured SAS images collected at sea. As the number of views of an object increases, classification performance improves and the dependence on the orientation of the object lessens.