Research on Obstacle Recognition for UUV Based on Multi-beam Forward Looking Sonar

Abstract

Obstacle recognition is one of the key technologies for Unmanned Undersea Vehicle (UUV). In this paper, the design of obstacle recognition based on multi-beam forward looking sonar is presented. First of all, Gaussian mixed model operator is adopted in adaptive threshold segmentation to segment the sonar image into regions of interest and background. Secondly, we use morphological  operation to remove noise and pseudo-target in the image. At last, obstacle contour is fitted into an ellipse using least squares method and then we calculate minimum and maximum distance, maximum and minimum azimuth angle between obstacle and multi-beam forward looking sonar. Results obtained on real sonar data are shown and discussed. Our experimental results show that the way can obtain accurate obstacle distance and azimuth angle information. DOI: http://dx.doi.org/10.11591/telkomnika.v11i5.2500