An Application of Pattern Recognition Techniques to Side Scan Sonar Data

This paper summarizes a study designed to determine if computer-aided pattern recognition techniques could be applied to the data of side scanning sonar. The technique presented is based on the hypothesis that variations in backscatter of sound energy from different ocean bottoms may be described in terms of changes in the texture and tone of side scanning sonar images. Measures of texture and tone are derived from first order statistics of the spatial variation of bottom returns. The pattern recognition classification algorithm known as "adaptive sample set constrution" is implemented for use with side scanning sonar data. The results are presented for application of the algorithm to the classification of two sand and one mud bottom type.