Automated mapping of marine habitats from marine sonar

Mapping marine habitats is a key theme of current research and a vital tool in marine resource management. The use of remote sensing in this field is now well established with both side-scan sonar and swathe bathymetry data being routinely utilised to map biological resources. Detailed interpretation of these data sets is a very time-consuming exercise and one that is often repeated by several different parties according to their field of expertise. This project investigated innovative ways/approaches of developing automated methods for the multi-use interpretation of marine geophysical data. It uses state-of-the-art techniques that have been developed for processing radar technology based image data for applications such as flood mapping and other terrestrial applications based on surface texturing. Such techniques are readily applicable to marine geophysical data, as has been confirmed through initial exploratory work. There is significant potential to support the expert data analysis through the application of a range of automated image processing techniques, which could accelerate processing as well as bring more rigor and consistency to the process. This project aimed to explore the potential of applying scientific expertise in image analysis in order to support the sonar and bathymetric data processing operation. More specifically, the research consisted of developing and comparing different automated algorithms to extract features of high scientific/conservation value as well as sedimentary boundaries from existing geophysical datasets. These algorithms are based on fusing expert knowledge and morphological image processing techniques commonly applied to terrestrial imagery from active radar sensors. The objective of this is to automatically filter raw acoustic data in order to define boundaries and identify features of interest, and validate these methods against existing ground truth data. Here, defining and using benchmark data sets would enable blind testing of the developed algorithm(s) and also produce freeware software tools implementing the most successful algorithms for the processing of raw sonar data onboard survey vessels, thus also allowing widespread use. Catalogues of sonar image processing techniques and the biophysical characteristics of seabed features of interest have been built as a way of constructing a project framework. This has highlighted a number of challenges and obstacles that need to be addressed. This is an exploratory project, designed to test the usefulness of novel methods for marine surveys. The project is not intended to be a complete evaluation for all substrate types; it is a test run to see if novel technologies are worth carrying forward to assist marine surveys.