On-Line High-Resolution Sonar Image Recognition Technique for Autonomous Underwater Vehicles

This paper addresses sonar image’s automatic object recognition method for AUVs (Autonomous Underwater Vehicles). The sonar images are severely changed depending on its poses. We propose active sonar image’s template updating method. Depending on the pose, a predictor modifies the template for the optimal recognition using the sonar’s camera model. This allows reliable recognition at any sonar pose. An experimental vehicle has been constructed to implement the proposed method and evaluate its capability in field.