SAS and Bathymetric Data Fusion for Improved Target Classification

SHORT ABSTRACT: An algorithm is proposed for the fusion of multiple views of an object from each of two information sources – a synthetic aperture sonar (SAS) image and a bathymetric map. A parameter tied to the success of interferometric processing, and hence the reliability of the bathymetric estimates, automatically weights the relative contribution of each information source. The variation in fused images, measured by the Laplacian, is used to determine the image translation needed to align multiple views. The algorithm is completely model-free and requires no a priori knowledge about the types of objects that will be considered. As a result, the method has potential to be particularly useful for reducing false alarms generated by clutter objects, and in turn, for improving classification performance. The proposed fusion algorithm is demonstrated on three objects using real, measured data collected at sea.