Sonar image processing: An application of template matching through relaxation

This paper dlscusses the use of high-level templates and relaxation techniques in processing side-scan sonar images. The images examined do not have enough resolution to identify the objects individually. Distances and orientations among groups of objects are stable, however, and provide the information necessary to unambiguously identify each object. The identified targets are then used to provide navigation information for underwater vehicles. Much of the work described is an application and refinement of established techniques, most directly those of Davis. Low-level filtering, template matching, relaxation and M. search are all discussed.