Synthesis and wavelet analysis of side-scan sonar sea bottom imagery

A backscattered side-scan sonar signal contains indirect information about the scattering surface, namely, the bottom sediment, character of bottom surface, and seafloor relief. This paper presents a method of automatic estimation of height of seafloor characteristic objects and micro relief reconstruction applying the measured shadow length. The proposed method utilises coefficients of two-dimensional discrete wavelet decomposition of a side-scan sonar image as the input to the self-organised neural network classification algorithm. The heights of seafloor characteristic objects were the basis for synthesis of a three-dimensional map of the bottom surface. The computations were conducted for data recorded in Hornsund Fjord (Spitsbergen Island, Svalbard Archipelago) during a habitat mapping experiment and for synthetic data. The verification of the proposed algorithm was made by comparison of computed results with calibrated video recordings