ENVIRONMENTALLY ADAPTIVE AUTOMATIC DETECTION OF LINEAR SEAFLOOR FEATURES IN SIDESCAN SONAR IMAGERY: THE CASE OF TRAWL MARKS DETECTION

This paper makes a detailed introduction to the design of a new automatic linear seafloor feature detection algorithm and its implementation on sidescan sonar (SSS) records, with its focus on trawl marks (TMs) detection. TMs are long linear scars on the seafloor, which are products of bottom trawl fishing. In image processing, detection of lines is a classical problem that seems to be more challenging on underwater acoustic imaging as the line segments are intermixed with wide-ranging environment backgrounds, and acoustic radiometric and geometric artifacts. Therefore, classic edge detection techniques based on the intensity gradient of the image do not yield reliable detection on SSS images. This proposed method integrates the characteristics of the linear features of interest in an environmentally adaptive procedure and is divided into three major steps. At the first step, preprocessing image techniques are applied to the original images. In the main stage, a spatial-domain filter is implemented through multi-scale rotated Haar-like features and integral images that measures the level of multiple oriented contrasts between adjacent areas. Seafloor characterization based on Anisotropy and Complexity calculations over the Haar-like filter’s responses identifies three types of seafloor texture: complex (e.g. biogenic mounds, clutter), anisotropic (e.g. TMs, ripples), and plain (e.g. undisturbed sand). At the same step, another function over filter’s responses produces a map that highlights the accurate locations where the candidate linear features prevail. The produced map is automatically binarized, morphologically processed and every linear image object is undergone properties measurement. The final linear features are selected according to a set of geometric and background textural feature criteria. In this study, is presented a set of assignment criteria that is tailored to the specific needs of TMs detection and is followed by TMs quantification that provides valuable measures for the estimation of bottom trawling impacts.