A Real-time Underwater Object Detection Algorithm for Multi-beam Forward Looking Sonar

A novel algorithm for the detection of underwater man-made objects in forward-looking sonar imagery is proposed. The algorithm takes advantage of the integral-image representation to quickly compute features, and progressively reduces the computational load by working on smaller portions of the image along the detection process phases. By adhering to the proposed scheme, real-time detection on sonar data onboard an autonomous vehicle is made possible. The proposed method does not require training data, as it dynamically takes into account environmental characteristics of the sensed sonar data. The proposed approach has been implemented and integrated into the software system of the Gemellina autonomous surface vehicle, and is able to run in real time. The validity of the proposed approach is demonstrated on real experiments carried out at sea with the Gemellina autonomous surface vehicle.