Rectilinear path detection of a target in the water column byusing a three-dimensional hough transform of forwardlooking sonar data

This paper presents a new approach to the problem of tracking objects in sequences of forward looking sonar images. Unlike previous work, we propose a “track before detect” procedure based on a specific Hough Transform. The Hough transform is a famous method for detecting lines in binary images but in this study, FLS sonar data are in grayscale and plotted in three dimensions. In typical FLS sonar images indeed, pixel intensity represents the backscattering at a given distance and a given azimuth but for a large range of elevations. The standard Hough transform has been adapted to this particular data set, and preliminary results have been obtained on real data.