Sonar and Optical Multimodal Detection of underwater objects

Roee Diamant and Tali Triebitz - Department of Marine Technologies

Computer vision
Image processing

Marine Science

Seed Grant 2020

Archeology usually brings to mind Indiana Jones style expeditions, but have you ever wondered how marine archeology is handled? Roee Diamant and Avi Abu from the Underwater Acoustic & Navigation Laboratory with the help of Tali Triebitz from the Marine Imaging laboratory, all part of the University of Haifa’s Data Science Research Center, are developing new data science ways to detect wrecks and sunken Neolithic settlements by a single autonomous underwater vehicle (AUV). For accurate object detection, the method combines the advantages of sonar images, which covers large areas, with those of optical images, which are of high resolution. The multimodal detection was made possible by developing a new type of feature descriptors that locks onto man-made objects, i.e., shapes of smooth contour. Their works [here, and here] showed great potential for feature analysis of sonar images, and their current results for merging real sonar and optical images show great potential.