Marine Sediment Mapping Using Multi-Source and Multi-Dimensional Acoustic Images Based on Evidential Fusion

Xi Chen,Jing Li,Wei Shen,Liangliang Tao,Yaokui Cui,Yang Hong

DOI: https://doi.org/10.1109/igarss.2018.8517470

2018-01-01

Abstract:This paper proposes a novel method to fuse multi-source acoustical remote sensing images for marine sediment mapping. Acoustic images from sidescan sonar, multibeam bathymetry, and sub-bottom profiler describe the 2-D, 2.5-D, and 3-D properties of the seabed sediment respectively. In attempt to make use of the multi-dimensional information from the multi-source data, the evidential fusion method, combined with object-based classification and spatial overlay analysis is proposed. Firstly, marine sediments are independently classified in the multi-source acoustic images with object based methods. Then, the spatial overlay analysis is conducted to group the classification results as evidence of different sediments. Finally, the evidential fusion method is employed to determine the exact distribution of sediments on the map. The proposed method introduces the classification results of sub-bottom profiler data that provide useful information, even though the data is limited in spatial coverage and is rarely used in automatic sediment mapping. The experiments show that the fused data from the three different sources of acoustic images significantly improve the mapping accuracy.