Sidescan sonar image processing for AUV navigation

The use of information extracted from payload data offers a potential improvement to the navigation of autonomous underwater vehicles (AUV). Recent research results have demonstrated the capability of bathymetric sensors and forward looking sonar to reduce navigation drifts of AUV's. As AUV's mostly perform data acquisition along straight lines, the side-scan sonar becomes a conventional payload for AUV's. Hence, the extraction and the processing of relevant information from the side-scan sonar images can be implemented on most AUV's as an aid to complete their mission. This aid applies at two levels of the mission control : improving the accuracy of the navigation and, at a higher level, validating the execution of mission segment by underwater landmark recognition. The approach developed here differs from concurrent mapping and localization in the sense that the knowledge of the area is required before the execution of a mission. The paper describes the creation of the reference database for navigation, the extraction of relevant information from the side-scan sonar images and the algorithms designed for matching current side-scan sonar observation with the reference data base. The results obtained in 2004 sea trials with GESMA's AUV REDERMOR2 are presented and discussed.