Target Detection in Clutter for Sonar Imagery

This proposal is a direct response to Challenge 22: Target detection in clutter. The Oceans Systems laboratory have a large experience in Target Detection and Classification in high resolution sonar imagery using a variety of techniques based on target responses and highlight shadow pairs[1,2,3]. Experiments on real scenarios show that these techniques work well in benign environments (flat and small ripples seabed) but deteriorate dramatically in complex scenarios[4]. Our experience with seabed classification based on 2D textures [5,6] also highlights that targets, even in complex cases, can appear as local outliers in the texture field. Examples of such scenarios and associated detections are shown below: