Conditioning of acoustic signals for improved sea floor characterization and imaging

摘要

A classification system was developed for seabed recognition using acoustic backscatter. The data was acquired by a Simrad EA 600 with two frequencies (38 - and 200-KHz). The algorithm was based on extracting raw data from just before and a fixed time after bottom detection. The resulting time series was Fourier transformed, and reduced to two features using PCA analysis. A gaussian classifier was used to assign the resulting features to classes. The resulting system seemed to perform satisfactory until a test involving sampling sediments at different heights were performed. The influence of depth, and several other "artefacts", on classification results is discussed.