Target Detection and Classification Capabilities of Two Multibeam Sonars

Emma Cotter, United States, Christopher Bassett, Applied Physics Laboratory, University of Washington, Seattle, WA, United States and Brian L Polagye, University of Washington Seattle Campus, Mechanical Engineering, Seattle, WA, United States
Multibeam sonars are used for a variety of oceanographic applications, ranging from detection of fauna in the water column to detection of gas seepages on the seafloor. In this study, we compare the target detection and classification capabilities of two multibeam sonars, the Tritech Gemini 720is and the Teledyne BlueView M900-2250. The sonars were deployed concurrently in a narrow tidal channel in Sequim Bay, WA with the widest dimension of the sonar swath parallel to the seafloor. The fields-of-view of the two multibeam sonars (120 x 20 degrees, range of 10 m) overlapped to the highest extent possible. Automatic target detection and tracking are used to isolate targets in the recorded data, and we compare the appearance of detected targets between the two sonars. We observe discrepancies in the appearance of concurrently detected targets due to the sonar frequencies and processing algorithms. Detected targets include diving birds, seals, schools of fish, and non-biological targets resulting from boat wakes and bubbles. A random forest algorithm automatically distinguishes between non-biological targets, fish schools, and individual biological targets, and we compare the automatic classification capabilities of each sonar. Finally, we discuss the implications of these findings and make recommendations for selection and configuration of multibeam sonars for target detection and classification.