An algorithm based on the wavelet transform for the classification of seabed textures

Despite the significance of the map of seabed textures for fishing grounds, geological survey using high resolution side-scan sonar is not widely utilized in fisheries due to its high cost. In order to use a standard fishing echo sounder instead of multi-beam echo sounders for the geological survey in classification of seabed textures, an algorithm based on the discrete wavelet transform is presented in this paper. The goal of the present algorithm is set to develop some kind of simple classification algorithm for fishery industry. Typical types of seabed textures such as sand, gravel, rock and reef were investigated and the values of wavelet powers at different levels of frequencies in wavelet analysis were compared to find characteristics of each reflected wave among different seabed textures. By the investigations, the authors found that the significant characteristics of each seabed texture can be seen especially in the lower frequency levels in the wavelet transform. In order to develop an automatic classification algorithm using a quantitative echo-sounder, the authors yielded linear discriminant functions for the automatic classification of seabed textures. Seventeen parameters were used in the discriminant functions including the integrated and maximum values of wavelet powers at different levels and statistics of the reflection waves (i.e. average, variance, maximum and minimum values). Consequently, sand and gravel seabed were well classified with more than 90% accuracies, and rock and reef seabed were classified with about 86% accuracies respectively. Based on the present algorithm, a software written in Visual Basic for MS-Windows was made and used for the demonstration of practical use of the present algorithm.