Riverbed Segmentation Using Image Processing Principles Applied to Multi-Beam Echo-Sounder Data

Seafloor mapping and classification are required in many underwater operations of which acoustic signals have been playing an important role. Multi-beam echo-sounders (MBES) are acoustic systems that provides full coverage high resolution bathymetry maps and backscatter information from the seabed or riverbed. They transmit a series of beams to the riverbed and receives the backscatter strengths (BS)، which depend on the angle of incidence. Angular dependence has already been applied to classify sediments using probability density function of the BS from MBES data. A complication occurs when the sediment types change along the swath، because it is difficult to separate from the beam angular dependence and the true backscatter strength variations. Therefore، the classification method is applied to the MBES data of specific angle. This paper employs an algorithm based on image processing principles، which the formation of the image from total massive data is the main step. After correcting for the angular dependence effects، they are gridded with an appropriate interpolation method. The image mosaics are then generated. The image processing techniques will then be applied to enhance the quality of images. Riverbed segmentation is performed by watershed transformation and morphological operations. Finally we assign a particular sediment type to each segment called ''labeling«by the mean grain size of the sediment grab sampling. The algorithm has been implemented on a MBES data set of the Waal river، the Netherlands. The correlation between the true data and image segmentation is about 63% using the specific segmentation. The results show that the proposed procedure improves the classification results compared to the traditional methods.