Underwater Object Detection with Efficient Shadow-Removal for Side Scan Sonar Images

Side scan sonar has been widely used in ocean investigations, underwater object detection by side scan sonar is one of the most essential and fundamental tasks these years. In this paper, we present one simplified underwater object detection scheme with the help of shadow removal of side scan sonar images. The fuzzy C-mean clustering (FCM) algorithm is first taken to partition all pixels from the side scan sonar images into a collection of C fuzzy clusters, which makes shadow regions be segmented. The Criminisi algorithm based on isophote-driven image sampling process is then made full use of to undertake the shadow removal by filling the shadow region. Through the Otsu algorithm by choosing threshold value automatically, to segment the object and background we complete the object detection. It is shown from the simulation experiments that the proposed approach could achieve great performances in the object detection with both robustness and effectiveness.