Automatic detection of underwater chain links using a forward-looking sonar

Abstract

Underwater chain cleaning and inspection tasks are costly and time consuming operations that must be performed periodically to guarantee the safety of the moorings. We propose a framework towards an efficient and cost-effective solution by using an autonomous underwater vehicle equipped with a forward-looking sonar. As a first step, we tackle the problem of individual chain link detection from the challenging forward-looking sonar data. To cope with occlusions and intensity variations due to viewpoint changes, the recognition problem is addressed as local pattern matching of the different link parts. We exploit the high frame-rate of the sonar to improve, by registration, the signal-to-noise ratio of the individual sonar frames and to cluster the local detections over time to increase robustness. Experiments with sonar images of a real chain are reported, showing a high percentage of correct link detections with good accuracy while potentially keeping real-time capabilities.