Multibeam echosounder data cleaning through a hierarchic adaptive and robust local surfacing

Multibeam echo sounders (MBES) datasets generally contain sporadic outlier points. The huge volumes of MBES datasets in a hydrographic framework require the use of semi-automatic techniques. In very shallow waters depth, data cleaning becomes a challenging task when potential dangers to navigation have to be carefully checked. The aim of our paper is to attempt this goal by combining two well-known techniques. The seafloor is constructed as an assemblage of surface elements with the help of a robust statistical approach. The local parameters model is a priori chosen, its scale is driven through a quadtree descending approach using subdivision rules based on both statistical and spatio-temporal inferences. Our multi resolution approach provides, with the algorithm outputs, a classification map that notes areas of concern.

► Our algorithm is dedicated to MBES data cleaning. ► It combines a robust estimator with a quadtree descending technique. ► The adaptive algorithm scheme uses MBES data in both geographical and time reference frames. ► Our algorithm provides a valid sounding dataset supplemented by a confidence level map. ► The performance of the algorithm is validated for artificial and real datasets.