Novel Sonar Technology

Novel Sonar Technology for New Discoveries of Undersea World

The Marine Technology Society (MTS) and the Oceanic Engineering Society of the Institute of Electrical and Electronics Engineers (IEEE) co-sponsor an annual conference on ocean sciences, engineering and technology, as well as policy. In addition to several hundred technical papers presented at the conference, papers with students as primary authors can be submitted for the Student Best Poster Competition. About 20 from among a large number of entries are selected to compete in the final competition. The student authors present their posters to half a dozen or more technical judges from the IEEE/MTS societies, assigned to score the entries based on technical contribution and effectiveness of presentation (in terms of style and clarity) in order to select the three best papers.

This year’s IEEE/MTS Oceans Conference was held in Monterey, California, September 19-22. In addition to two technical papers presented by the UM researchers at the conference, the paper entitled “Segmentation, Classification and Modeling of Two-Dimensional Forward-Scan Sonar Imagery for Efficient Coding and Synthesis” was selected for the Best Student Poster Competition. Co-authored by the team of Ph.D. student Mohammad Haghighat, two undergraduate students (Xiuying Li and Niching Fang), visiting scholar Yang Zhang and Professor Shahriar Negahdaripour, the paper won the second-place award at the conference.

One aspect of Professor Negahdaripour’s research deals with developing computer algorithms for the automated interpretation of two-dimensional underwater sonar images, acquired under poor visibility conditions, which prohibit the use of optical images. A key contribution to oceanic engineering is to facilitate a wide variety of underwater tasks and operations — carried out by autonomous robotics vehicles — that are critical for new discoveries of the undersea world, and to today’s global economy (e.g., fish stock assessment, seafloor and habitat mapping, and inspections of underwater pipelines and other structures). In this work, computer algorithms segment and classify the sonar images into objects and their shadow areas as signal components and the less informative background region. This provides a more compact image representation that can be transmitted more efficiently through low bandwidth acoustic channels to a surface station, allowing effective two-way interaction with the submersible robot in real-time.