Fast trend extraction and identification of spikes in bathymetric data

Average interpolating subdivision is a fast method for finding trends in data sets collected with multibeam echo sounders. By grouping a suitable number of pings together, a swath is sectioned into a series of geographical areas that are analyzed consecutively. This makes it possible to exploit both cross-track and along-track correlations in identifying outliers. A one-dimensional example shows how the algorithm behaves in the presence of spikes, burst noise, and random measurement errors. More realistic experiments are carried out on multibeam echo sounder data collected with an EM 1002 system. The algorithm is fast enough for real time identification of spikes.