Target Detection and Parameter Recognition for The Crawling Submersible Based on Forward-looking Sonar

The crawling submersible is a new type of deep-sea unmanned submersible that can both cruise in the deep sea and crawl on the seabed. The forward-looking sonar is used by the submersible for obstacle avoidance in the crawling process. This paper proposes a new target detection and parameter recognition method using multi-frame acoustic images from the single-beam and mechanical scanning sonar. The proposed target detection and parameter recognition method implements first the multi-frame acoustic image synthesis and filtering algorithm for overcoming the “overflow” problem after fusing the images, next uses the piecewise linear stretch and median filtering algorithm to deal with the blurring issue at the target edge and to reduce the noise interference, adopts the Otsu algorithm for determining the optimal segmentation threshold of the binarization operation, implements the mathematical morphological operation to further improve the quality of the target detection, and finally detects the target and obtains their location and distance information through searching for the connected domain in the images. The experiment results demonstrate that the proposed target detection and parameter recognition method is able to reduce the occurrence of false alarms and effectively obtain the position and distance information of targets relative to the crawling submersible. This will provide the basic information for the crawling submersible to achieve efficient obstacle avoidances.