An Overview of Multibeam Sonar Data Processing for Benthic Habitat Mapping at Deakin University
Ierodiaconou, D and Schimel, A and Rattray, A and Monk, J and Che Hasan, R and Blake, S, An Overview of Multibeam Sonar Data Processing for Benthic Habitat Mapping at Deakin University, Workshop on Acoustics and Automated Video Processing for Fisheries and Environmental Monitoring, June, Hobart, Tasmania (2013) [Conference Extract]
PDF (An Overview of Multibeam Sonar Data Processing for Benthic Habitat Mapping at Deakin University. ) Pending copyright assessment - Request a copy 13Mb
Deakin University has been carrying out research in habitat mapping techniques applied to the coastal regional Victoria since 2007, using primarily bathymetry and backscatter data from Multibeam sonar, coupled with video ground-truth. Contrary to the traditional seafloor-mapping approach of backscatter imagery segmentation, our research has been focused on the implementation of ecological modeling approaches, which consist in letting ground-truth data drive a modeling algorithm using a varied set of "predictor" data layers that include –but are not limited to - bathymetry and backscatter data from Multibeam sonars. Slight modifications can be implemented to these approaches to produce two types of outputs presenting different ecological interests: "benthoscapes" (maps of seafloor bio-physical properties) and "habitat suitability maps" (habitat preference maps for single species). Such ecological approach to mapping the seafloor presents the main advantage to allow for any geographical dataset and its derivatives to be implemented and tested for its predictive power. Tested variables to-date include Multibeam and Lidar bathymetry data and derivatives (slope, rugosity, etc.), Multibeam backscatter data and derivatives (HSI, textural features, 1st-order statistics, etc.), but also backscatter angular response features, hindcast oceanographic conditions, etc. Our most recent work has been focusing on (1) processing backscatter angular response to create additional layers of information, (2) implementing random-forest as classification algorithm to allow for the estimation of the relative importance of each variable, (3) developing procedures to ensure the repeatability of our studies, that is, the reliability of results, and allow for detection of changes over time. Deakin University recently launched its own research vessel MV Yolla and equipped it with a Kongbserg EM2040c Multibeam sonar aided with a POS MV positioning/attitude system. Future research directions include exploring the new capabilities of this next-generation multibeam sonar in terms of water-column imaging and wide bandwidth, in order to derive new data layers for our classification processes. The lack of calibration of Multibeam sonar backscatter remains a major obstacle, especially affecting our development of repeatable procedures across different systems and study sites.