The Oceanographic and Hydrographic service of the Navy (SHOM) has been using two MultiBeam Echo-Sounders (MBES) since 1988. These systems enable swath coverage of the sea floor along a survey line.
Compared with single beam Echo-sounder systems, the resolution of the data provided by these systems has been considerably increased. Nevertheless, errors still remain and they must be detected and eliminated to meet the international standards of bathymetric charts.
The high volume of data, particularly in the case of very shallow water Echo-Sounder systems, makes manual validation of the data inappropriate. In order to reduce the operating costs of the data cleaning step, SHOM has developed algorithms to automatically detect huge datasets generated by MultiBeams.
The algorithm described in this paper is based on a local modelization of the seabed. The fitting of a quadratic surface over the raw data is carried out using a robust estimator. We retained Tukey robust estimator as the most effective choice due to its adaptative capabilities. Possible outliers are soundings with high residual values between measured depths and depths estimated from the local model. Retained outliers are deduced from this first outliers set, by computing local cross validation.
This algorithm has been tested on different bathymetric data sets. Its efficiency has been demonstrated whatever the depth or type of seabed. Moreover, its application only requires two parameters to be set, thus making it the obvious choice. It has currently been adopted and installed on board all the SHOM's ships.