Automatic Sea Floor Characterization based on Underwater Acoustic Image Processing

Automatic sea floor characterization is mainly based on the signal or image processing of the data acquired using an active acoustic system called sediment sonar. Each processing method suits a specific type of sonar, such as the monobeam, the multibeam, or the side-scan sonar. Most types of sonar offer a two dimensional view of the sea floor surface. Therefore, a high resolution image results which can be further analyzed. The inconvenient is that the sonar cannot view inside of the sea floor for a deeper analysis. Therefore, lower frequency acoustic systems are used for in-depth sea floor penetration (boomer, sparker, airguns or sub-bottom profilers). In this case, a mono dimensional signal results. Previous studies on the low-frequency systems are mainly based on the visual inspection by a geological human expert. To automatize this process, we propose the use of feature sets based on the transposed expert fuzzy reasoning. Two features are extracted, the first based on the sea floor contour and the second based on the sub-bottom sediment texture. Key–Words: Sedimentology, Underwater Acoustics, Pattern Recognition, Image Processing, Textures, Wavelets