High-resolution Delineation of Acoustically Homogeneous Areas in Multibeam Backscatter Maps


A complete description of essential seafloor habitats requires the analysis of multiple properties acquired from the seafloor, the water column, and the surface properties. Nevertheless, one essential step in this analysis is the characterization of the seafloor substrates. There were attempts to accomplish this task through optical techniques, which offer the ability to provide very high-resolution (mm-scale) images of the seafloor, but the limited propagation of light in seawater severely restricts the seafloor area that can be covered. A more viable option are acoustic techniques, mainly multibeam echosounders (MBES) , which offer the ability to cover broad areas with high spatial resolution and thus the ability to provide detailed maps of morphology and rugosity. Calibrated acoustic backscatter acquired by multibeam sonars can be used to provide estimates of the grain size and impedance (the product of sound speed and density) of the sea floor through the inversion of acoustic backscatter model that takes into account the angular dependence of backscatter (ARA – Fonseca and Mayer, 2007). The use of angular dependence for seafloor characterization is normally limited to “patches” of the seafloor that have dimensions on the order of the swath width of the sonar, thus constraining the spatial resolution of the characterization process. In order to surpass this limitation a “thematic” approach was developed to analyze seafloor backscatter data – however the thematic approach requires an a priori delineation of areas of the seafloor with common acoustic response. We present a new automatic approach for delineation of areas representing different acoustic facies. The technique is not directly dependent on the mosaic, the appearance of which is affected by the particular method of its construction (the mosaic is essentially an example of many-to-one mapping), and utilizes all the backscatter data available for the survey. Automatically determined boundaries derived using this approach correlate well with those obtained by manual procedures and available groundtruth data yet have much higher resolution. Thus, the new approach provides a rapid, unbiased and objective process for the high-resolution delineation of regions of the seafloor of common acoustic response which can then be input into an appropriate inversion model for seafloor characterization. In recent years, many attempts have been made to develop automatic methods for segmentation of hydro-acoustic remote sensing data acquired by multibeam echosounders in order to generate quantitative estimates of the spatial distribution of seafloor relief, bottom type and composition. The majority of the segmentation methods presented so far have been based on image processing techniques, which assume implicitly that the backscatter mosaic reflects all the available backscatter data. This limits their ability to unambiguously discriminate seafloor properties, as the primary observation of an MBES is not backscatter imagery or mosaics, but rather backscatter angular response. Mosaics are only projections of the original observations, with resulting loss of information. The method we are developing is fully automatic and attempts to segment the acoustic remote sensing data simultaneously in the image-textural space and in the angular-response space. The output of this automatic procedure is a thematic map, where the individual themes have boundaries defined at the mosaic image resolution, but still have sufficient angular coverage to allow for seafloor characterization. Angular Range Analysis (ARA) inversion is then applied to the average angular response of individual themes, generating estimates of the acoustic impedance, acoustic roughness, and mean grain size of the seafloor within each theme. The technique described above is applied to a Simrad EM1002 95kHz MBES dataset acquired from a study area covering an offshore reef at Stanton Banks, UK. The results are compared to still photo images, grab samples and previous habitat maps existent in the area, to asses the ability of the acoustic theme segmentation to discriminate benthic habitats.

Publication Date


Journal or Conference Title

Gulf of Maine Symposium

Conference Date

Oct 4 - Oct 9, 2009

Publisher Place

St. Andrews, New Brunswick, Canada


Fisheries and Oceans Canada

Document Type

Conference Proceeding

This document is currently not available here.