Comparative study of shallow water multibeam imagery for cleaning bathymetry sounding errors

Abstract

Presents the results of a six-month study for the French Hydrographic Service (SHOM) to investigate the use of multibeam seafloor imagery for aiding existing bathymetry data cleaning techniques. These data cleaning algorithms efficiently eliminate erroneous soundings from deep water (depth >80 m) survey datasets but generate dubious soundings in shallow water. Such soundings are time consuming for an operator to validate or invalidate. In order to improve performance, the authors tested whether additional information could be derived from the correlation between multibeam imagery and bathymetry. The discussed methodology attempts to associate imaged objects (echo/shadow sets) with a list of suspicious soundings output by SHOM algorithms. Two approaches are considered: a ping-to-ping approach and a geographic approach. Object detection algorithms are run on the two different methods. Two datasets are examined: one from a SIMRAD EM1002S and another from an ATLAS FS20. The segmentation tools developed are helpful for analyzing suspicious beams where the imagery presents an anomaly. The four methods implemented may be adapted to the type of data used and to the desired subtlety of the segmentation.