Sonar image interpretation and modelling

Abstract

The problem addressed is that of providing a model-based system to interpret sonar data from an autonomous or remotely controlled vehicle instrumentation, to aid navigation, improve sea floor mapping techniques, identify objects, simulate acoustic images for survey and analysis, or to aid the design of sonar systems. Analytic tools have been developed to segment and classify sea-bed and shallow seismic image textures, remove multiple reflections and detect objects on or close to the sea-bed. The sonar interpretation problem is placed within a model-based framework. Three dimensional surfaces are modelled to represent accurately natural seabed and sedimentary layers, so that the model parameters can provide a confident measure of the real data characteristics. Realistic synthetic sonar images are generated using the model-based simulation. In addition, new analytic tools can be developed and evaluated in a controlled experimental environment. This paper presents the conceptual sonar simulation system architecture, outlines the visualisation of the 3D sea floor model, describes the sonar system simulation and references image analysis tools available for texture segmentation and object detection.