Underwater target detection and classification using high reso- lution sensors and wide band sonar

In this presentation, our work on integrated target detection and identification using sonar systems is presented. Standard high‐resolution imagingsonar such as side scan and synthetic aperture are used for detection. Existing methods focus on statistical techniques using Markov models and machine learning‐based techniques. Results on both synthetic and real data acquired in a large variety of terrains are presented. Identification can be based on either the shape of the target or its internal structure. Techniques‐based active contours for shape analysis have been developed but these techniques only work well on very high‐resolution data such as SAS and can be difficult to use on cluttered terrain. We propose to use the wide‐band bio‐sonar we developed to perform this task. In that case, the analysis focuses on the resonances induced by the internal structure of the target and the material used. Simulation confirmed by theoretical analysis on simple targets and experimental results demonstrate the effectiveness of this approach. These techniques are integrated on real vehicles in a collaborative vehicle framework to perform detection, classification, and identification of targets. [This work was supported by Thales, The Ministry of Defence, and the European Union under contract MRTN/CT/2006/036186.]