Underwater target´s size/shape dynamic analysis for fast target recognition using sonar images

Underwater target recognition techniques are very important for an autonomous underwater vehicle (AUV) to complete its missions such as relocation of lost targets, removal of toxic wastes and mine counter measures, etc. In this paper we address the problem of how to extract the target´s size/shape features from a sequence of 2D acoustic images acquired by a forward looking sonar (SeaBat 6012). We make use of image processing techniques and the near field acoustic scattering principle of underwater target to estimate the two-dimensional position and size/shape of a nearby target. These features are very useful for an AUV to design its obstacle avoidance or recognition strategies. Some experimental and analytical results about these features are provided to demonstrate the feasibility of these methods