Order Statistics Filtering for Detecting Outliers in Depth Data along a Sounding Line

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

“…The performances of all the algorithms with the exception of those described in [39,47] were measured on MBES datasets. Li and Sedaghat data-oriented approaches were tested on SBES datasets.…”

“…There can also be a combination of the three previous methods. Data-oriented algorithms are most commonly assessed using visual inspection method (i.e., 7 approaches [2,30,35,38,39,46,50]). Among the remaining, five [4,16,17,31,32] of them rely on the DBM method and only two apply a combined validation method [27,33].…”

“…• Isolated outliers (Figure 5b); • Structured groups of outliers (Figure 5a,c,d); • Unstructured groups of outliers (Figure 5d); Some authors mention and propose to detect only one type of error, namely isolated outliers for Shaw [34], spikes for Eeg [35], Ferreira [36] and Bjorke [37] and impulsive noise for Mann [38]. Calder, Du, Li, Lu and Motao [17,27,[39][40][41] oppose this first type of outliers to a second type consisting of groups or clusters of outliers or burst data, emphasizing the difficulty of dealing with them. Isolated pings or defect beams fall into this category.…”

“…Significant slope occurring across the swath of the sonar generates inconsistent beams which are spatially coherent and shoaler than the seafloor. Such configurations of outliers are particularly problematic for experts and automatic procedures leading to asymmetric processing in a hydrographic context, by processing soundings detected as erroneous differently depending on whether they are located above or below the surrounding seabed [25,39,41].…”