Autonomous processing of sidescan sonar imagery for unmanned underwater vehicles

Unmanned Underwater Vehicles can autonomously obtain high quality sidescan sonar backscatter swathe imagery of the seabed by traversing pre-programmed tracks at a set height above the bottom where they are relatively undisturbed by wave action. Some survey activities could be more efficient if the UUVs had some machine intelligence or decision making capability. For example, UUVs could be programmed to survey seabeds of particular interest at slower speeds to obtain more detail, or might resurvey particular areas using more detailed search patterns. Onboard processing for this purpose requires fast algorithms and robust decisions. Ongoing work in autonomously characterizing sidescan sonar imagery for seabed type is described. Humans can readily recognize patterns and textures indicative of particular seabed types, for example, sand has characteristic ripple patterns. Can Unmanned Underwater Vehicles be given the capability to autonomously mimic human visual perception?