Classifying the seafloor using a textural segmentation approach

Abstract

Seabed habitat mapping using multi-beam echo-sounder data is a very active field of research with direct uses in protecting ecologically important areas, marine resource management, and to set legislation to safeguard the oceans. For seafloor classification, it is important to use the best data possible but it is also important to extract the most information from the available data. Seafloor classification is often done by directly using backscatter, bathymetry, and bathymetric derivative data produced by multi-beam echo-sounder systems. A way to extract more information from the above-mentioned data is to also use texture information from the bathymetry and/or backscatter. In this study, texture based classification was performed on bathymetry data from the Borkumse Stenen and Bruine Bank within the Dutch sector of the North Sea. The method makes use of object-based image analysis (OBIA; using eCognition). The classification results are verified by using grab samples from the DINOloket database. The performance of texture based classification will be examined when bathymetry data alone is used as input. It will be further investigated if including texture based in conjunction with backscatter, and bathymetry based classification improves classification performance of currently existing methods. It will also be examined if rule sets developed for one area of the sea can be used to classify the seafloor in another area and the effect that differing spatial resolutions of different datasets have on the portability of texture-based classification rule sets.