A Review of Data Cleaning Approaches in a Hydrographic Framework with a Focus on Bathymetric Multibeam Echosounder Datasets

Julian Le Deunf,Romain Billot,Thierry Schmitt,Nathalie Debese

doi:10.3390/geosciences10070254

Abstract

Automatic cleaning of MultiBeam EchoSounder (MBES) bathymetric datasets is a critical issue in data processing especially with the objective of nautical charting. A number of approaches have already been investigated in order to provide solution in views of operationally reaching this still challenging problem. This paper aims at providing a comprehensive and structured overview of existing contributions in the literature. For this purpose, a taxonomy is proposed to categorize the whole set of automatic and semi-automatic methods addressing MBES data cleaning. The non-supervised algorithms that compose the majority of the methods developed in the hydrographic field, are mainly described according to both the features of the bathymetric data and the type of outliers to detect. Based on this detailed review, past and future developments are discussed in light of both implementation and test on datasets and metrics used for performances assessment.

Highlights

More and more bathymetric information is being acquired, motivated by the evolution of the sensors, the democratization of shallow water sensors through simplified system integration, citizen science or from global initiatives such as the EuropeanMarine Observation and Data network (EMODnet) Bathymetry, the Seabed 2030 General BathymetricChart of the Oceans (GEBCO) or the United Nations (UN) Decade of Ocean Sciences for SustainableDevelopment
Experience shows that the collected information contains sparse erroneous soundings that have to be invalidated before delivering digital bathymetric models (DBM)
The development of automatic approaches increased in the beginning of the late 90s to early 2000s with the use of shallow water MultiBeam EchoSounder (MBES) systems

Summary

Introduction

More and more bathymetric information is being acquired, motivated by the evolution of the sensors, the democratization of shallow water sensors through simplified system integration, citizen science (with crowdsourced bathymetry projects) or from global initiatives such as the EuropeanMarine Observation and Data network (EMODnet) Bathymetry, the Seabed 2030 General BathymetricChart of the Oceans (GEBCO) or the United Nations (UN) Decade of Ocean Sciences for SustainableDevelopment. More and more bathymetric information is being acquired, motivated by the evolution of the sensors, the democratization of shallow water sensors through simplified system integration, citizen science (with crowdsourced bathymetry projects) or from global initiatives such as the European. Marine Observation and Data network (EMODnet) Bathymetry, the Seabed 2030 General Bathymetric. Alongside this increase of bathymetric information, data processing is a critical step in order to ensure the best quality for multiple applications related principally to safety of navigation for all sea users and digital terrain modeling. Experience shows that the collected information contains sparse erroneous soundings that have to be invalidated before delivering digital bathymetric models (DBM). The detection and cleaning of erroneous soundings is critical, as the final objective is to plot bathymetric information on nautical charts, which have a legal status.

Objectives

Findings

Discussion

Conclusion