A method for multibeam seafloor terrain classification based on self-adaptive geographic classification unit. (1st January 2020)

• Record Type: Journal Article
• Authors: Cui, Xiaodong; Xing, Zhe; Yang, Fanlin; Fan, Miao; Ma, Yong; Sun, Yi
• Is Part Of: Applied acoustics. Volume 157(2020)
• Publication Date: 2020-01-01
• Subjects: Multi-beam echo sounding -- Seafloor terrain classification -- Terrain feature point -- Seafloor topographic recognition unit; Acoustical engineering -- Periodicals; Periodicals; 620.2
• DOI: 10.1016/j.apacoust.2019.107029
• Languages: English
• ISSNs: 0003-682X

Abstract

Abstract: Multi-beam echo sounding (MBES) provides high-precision and high-resolution bathymetry data that lay the data foundation for the refinement and accuracy of seafloor terrain classifications. Based on the digital depth model (DDM) of multibeam bathymetry, this paper proposes a seafloor terrain classification method using morphology and pattern recognition. First, a two-dimensional discrete wavelet transform (DWT) and self-adaptive threshold algorithm is applied to extract seafloor terrain feature points from the DDM. Then, a self-adaptive terrain element classification and mapping method is implemented based on pattern recognition. Its core is automatic determination of the geographic classification unit centered on the terrain feature point that including terrain type and unit area, thus the seafloor terrain is divided into 7 types of seafloor hill, depression, ridge, valley, shoulder, footslope and plain according the principle of similarity and multi-modality. For different resolutions of the DDMs, this method can automatically identify its spatial scale and determine the terrain recognition unit by a single scan of the seafloor feature points instead of all grid units. This method is applied to the multibeam bathymetry data collected in the middle of the South China Sea and automatically generates a seafloor classification map with an overall accuracy of 98.3% and Kappa of 97.6%.