Validation of automated supervised segmentation of multibeam backscatter data from the Chatham Rise

Hillman, Jess, Lamarche, Geoffroy, Pallentin, Arne, Pecher, Ingo A., Gorman , Andrew R. and Schneider, Jens (2017) Validation of automated supervised segmentation of multibeam backscatter data from the Chatham Rise Marine Geophysical Research . DOI 10.1007/s11001-016-9297-9 .

[img] Text
Hillman.pdf - Reprinted Version
Restricted to Registered users only

Download (11Mb)

Supplementary data:


Using automated supervised segmentation of multibeam backscatter data to delineate seafloor substrates is a relatively novel technique. Low-frequency multibeam echosounders (MBES), such as the 12-kHz EM120, present particular difficulties since the signal can penetrate several metres into the seafloor, depending on substrate type. We present a case study illustrating how a non-targeted dataset may be used to derive information from multibeam backscatter data regarding distribution of substrate types. The results allow us to assess limitations associated with low frequency MBES where sub-bottom layering is present, and test the accuracy of automated supervised segmentation performed using SonarScope® software. This is done through comparison of predicted and observed substrate from backscatter facies-derived classes and substrate data, reinforced using quantitative statistical analysis based on a confusion matrix. We use sediment samples, video transects and sub-bottom profiles acquired on the Chatham Rise, east of New Zealand. Inferences on the substrate types are made using the Generic Seafloor Acoustic Backscatter (GSAB) model, and the extents of the backscatter classes are delineated by automated supervised segmentation. Correlating substrate data to backscatter classes revealed that backscatter amplitude may correspond to lithologies up to 4 m below the seafloor. Our results emphasise several issues related to substrate characterisation using backscatter classification, primarily because the GSAB model does not only relate to grain size and roughness properties of substrate, but also accounts for other parameters that influence backscatter. Better understanding these limitations allows us to derive first-order interpretations of sediment properties from automated supervised segmentation.

Document Type: Article
Keywords: Bathymetry ; Backscatter ; Geomorphology ; Substrate classification ; Angular response ; Segmentation, New Zealand
Research affiliation: OceanRep > The Future Ocean - Cluster of Excellence
Kiel University
Refereed: Yes
DOI etc.: 10.1007/s11001-016-9297-9
ISSN: 0025-3235
Projects: Future Ocean
Date Deposited: 12 Dec 2016 09:41
Last Modified: 06 Feb 2019 15:45

Actions (login required)

View Item View Item

Document Downloads

More statistics for this item...