An Approach to Automatic Detection of Outliers in Multibeam Echo Sounding Data
An approach to automatic detection of outliers in multibeam echo sounding data is developed by simulating an operator's decision-making in interactive editing. Both data-oriented and space-oriented approaches are used to achieve flexibility in defining the working window. Clustering by mode seeking, termed unsupervised learning in pattern recognition, and the Dixon-type discordancy test based on a uniform probability density function are utilized to achieve robustness in detecting outliers. The approach is tested versus interactive editing by using real data collected at an area of known ground truth. It is concluded that the approach matches the results of interactive editing at 0.95 in terms of Rand index, no significant distortion in the bathymetry is found, and the processing time is decreased by a factor of 30.
Center for Coastal and Ocean Mapping Affiliate
the Hydrographic Journal
The Hydrographic Society
Du, Z, Wells, D. and Mayer, L.A., 1996, An Approach to Automatic Detection of Outliers in Multibeam Echo Sounding Data, The Hydrographic Journal, no. 79, pp. 19-23.