Image processing techniques for the analysis of sidescan sonar survey data

Sidescan sonar surveys are routinely carried out, for example, within the offshore oil and gas industry to gather information from the seabed. This paper presents several techniques which have been developed to analyse the data generated from such surveys. Classification algorithms are described which have been developed to perform automatic sediment identification and seabed mapping and the problem of inferring the presence, or otherwise, of objects on the seabed from the local backscatter is also addressed. A unified approach to object detection and seabed classification is adopted. The object detection algorithm which is presented exploits a knowledge of the local backscatter characteristics (as provided by the classifier) to enhance detection capability. A fast implementation of the classifier algorithm is also described which exploits the block-Toeplitz structure of the covariance matrix for each sediment class. This is important since large quantities of data are routinely gathered during survey operations for processing at a later date.