Seabed classification through multifractalanalysis of sidescan sonar imagery

The authors present a technique for the classification and analysis of seabed sediments from sidescan sonar images, the origins of which lie in the body of fractal theory. Six seabed types were analysed, namely clay, mud, sand, gravel, stones and rock. These data sets have previously been analysed by several authors who have used techniques based on the power spectrum. The method proposed in the paper allows frequency information to be obtained but without the use of large windows which are generally required for frequency domain measurements. Results are presented for the classification of individual ground truthed sediments and the segmentation of composite images containing these sediments. Correct classification rates of greater then 99% have been obtained and good segmentation accuracy achieved.