Detection and Elimination of Bathymetric Outliers in Multibeam Echosounder System Based on Robust Multi - quadric Method and Median Parameter Model

Abstract

Multibeam echosounder system is a dynamic measurement under the continuous motion of measurement platform. Contaminated sporadic outliers are inevitably generated during bathymetric data acquisition due to the interference of underwater environmental effects, such as ocean waves, wind, and tides. A filtering model of robust multi-quadric method based on median parameter was proposed in this study to detect and eliminate the outliers in bathymetric datasets. The submarine topography trend surface model was constructed by the multi-quadric function. Known node parameters in the function were obtained by the threshold extraction method of joining angle, chord with dispersion. The robust iterations of residual error initials and the weight-weakening process of outliers were studied by median parameters and IGGIII equivalent weight function. The validity of the proposed filtering model was verified by experimental datasets of offshore waters, which were collected by BV5000 multibeam side scan sonar. Results demonstrate that the known node selection strategy in this study can remove redundancy information in point set and extract regional characteristic points to the maximum extent. The fitting errors at normal bathymetric points are all below 0.06 m, which are decreased compared with the corresponding magnitude of other schemes (0.16 m, 0.12 m). The residual initial value of robust iteration based on median parameters exhibites relatively high estimation accuracy and detection efficiency. The errors of bathymetric points are relatively small (0.019 m), and the detection number of outliers (150) is efficient in the entire detection region. This work can improve the accuracy and efficiency of underwater terrain detection and lay a foundation for later scientific research and marine exploration.

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

“…Compared to a land survey, the outlier rate is higher in a bathymetric survey due to the complexity of the data acquisition process. In the bathymetric context, the outlier percentage is considered to be less than 1% for some authors [24,25], for others it is less than 10% [26], but can reach up to 25% for some tests [27]. The high variability of outlier percentage observed in the literature can be explained by the wide panel of sensors used as well as the high variability of the environmental conditions.…”

“…However, native SVM approaches are known to have a high computational cost. Wang ( [26,52]) proposes heuristics algorithms aiming at working through a reduced set of support vector machines, still using radial basis kernels and robust metrics.…”

“…This outlierness measure gives rise to a ranked list of detected soundings requiring the intervention of a hydrograph to decide whether a doubtful sounding must be invalidated or not. The content of the list may be fully observed pulling out the soundings in decreasing order, or, partially observed, applying an outlierness threshold [3,26,36,41,50]. This first approach was historically preferred as hydrographic services are responsible for navigational security [2].…”

Outlier detection for Multibeam echo sounder (MBES) data: from past to present

“…An improved version of CUBE is proposed with CHRT (CUBE with Hierarchical Resolution Techniques) including the multiresolution, multi-processing and taking into account the quality factor developed by Ifremer (Institut français de recherche pour l'exploitation de la mer) which could resolve the issues explained above. 2) RMQMP(2018) method The Robust Multi-quadric Method and Median Parameter Model (RMQMP) is described in [8] and in Fig .7. At first a fitting trend surface model is built, a median parameter method is used to obtain a first value of residual error which is applied as an initial value within an iterative process to weaken soundings' weights (considered as outliers) in the DBM generation.…”

“…Bathymetry Outlier Detection Procedure of RMQMP method from[8] Taken from[9], clustering by mode seeking.…”