Extracting seafloor elevations from side-scan sonar imagery for SLAM data association

Data association is a critical component of simultaneous localization and mapping (SLAM). This is challenging in an underwater environment with an autonomous underwater vehicle (AUV) where currents can alter the AUV´s perceived location of landmarks used to update the AUV´s estimated position. In an effort to reduce false positives in the data association seafloor elevation trends local to SLAM landmarks are used as additional features to assist in verifying associations between landmarks. Elevation gradients are less sensitive to sensor error and seafloor changes over time than other environmental features. Elevations are extracted from side-scan sonar data and new landmark elevation profiles are compared to previously observed ones to find the best associations. This paper reports on a unique ability to identify the best match within a set of landmarks and is a good complementary feature to an existing data association algorithm.