An approach to robust map generation from multibeam bathymetric data
During the last twenty years, many multibeam bathymetric sonars have been produced. The instrumentation is usually accompanied by a system able to produce a seafloor map from the sonar data. There are also several public domain systems, which can be used to obtain a map from the data. All these systems produce a gridded map that must be filtered in order to reproduce the original seafloor surface because of noise on the bathymetric data. Both the gridding and the filteringalgorithms introduce a source of error that is not easily controlled. Moreover, gridded maps may usesignificant storage space for a small amount of information. Finally, at present no systematic solution with realistic run-time requirements has been given to the problem of identification and elimination of bad data (outliers). We present here an algorithm able to fit bathymetric data and to automatically deal with outliers. The most important characteristics of the algorithm are: production of a triangulated map of uniform accuracy irrespective of seafloor features; a map resolution which depends on the local data noise amplitude; automatic elimination of outliers and low computing cost even on large data files. The algorithm can be used to reduce the operator intervention during bathymetric datamapping. Raw bathymetric data are directly analyzed by the algorithm which automatically and robustly eliminates outliers and produces a map the parameters of which can be finely tuned by the user (number of nodes, smoothing, level of data cleaning, etc.). The algorithm has been implemented, exhaustively tested on synthetic and real data and fully documented.