Error in target-based georeferencing and registration in terrestrial laser scanning

Abstract

Terrestrial laser scanning (TLS) has been used widely for various applications, such as measurement of movement caused by natural hazards and Earth surface processes. In TLS surveying, registration and georeferencing are two essential steps, and their accuracy often determines the usefulness of TLS surveys. So far, evaluation of registration and georeferencing errors has been based on statistics obtained from the data processing software provided by scanner manufacturers. This paper demonstrates that these statistics are incompetent measures of the actual registration and georeferencing errors in TLS data and, thus, should no longer be used in practice. To seek a suitable replacement, an investigation of the spatial pattern and the magnitude of the actual registration and georeferencing errors in TLS data points was undertaken. This led to the development of a quantitative means of estimating the registrationor georeferencing-induced positional error in point clouds. The solutions proposed can aid in the planning of TLS surveys where a minimum accuracy requirement is known, and are of use for subsequent analysis of the uncertainty in TLS datasets.

Long‐range terrestrial laser scanning for geomorphological change detection in alpine terrain – handling uncertainties

“…The registration error resulting from the registration of scans from different scan positions and from the multi‐temporal registration is not spatially uniform and can vary significantly between the central parts of the data set and surfaces deviating from the main slope direction. The calculation of the LOD could be further improved by the estimation of a spatially distributed registration error as proposed by Fan et al (), although this is difficult in this case.…”

“…An assessment of the spatial distribution of the registration error is hardly possible because the registration error cannot be distinguished from the positional uncertainties and the surface roughness. The assessment of the registration error by using the error statistics describing how well target constraints are matched (Oppikofer et al , ; Barnhart and Crosby, ; Lague et al ) has been shown by Fan et al () to be an incompetent measure, especially at larger ranges. However, these error statistics are often used for the estimation of the registration error and for uncertainty analyses.…”

Using terrestrial LiDAR to measure water erosion on stony plots under simulated rainfall

“…However, the density and arrangement of targets will affect this assessment and, moreover, this method presumes that the target center can be identified exactly by the scanner, which is not always guaranteed (Fan et al, 2015;Fey and Wichmann, 2017). However, it is challenging to estimate registration error because it is difficult to distinguish error associated with the mismatch of points between the two point clouds from scanning noise, which is a property of individual scans.…”

“…An overview of the evolution of soil surface roughness can be found in Li et al (2019). Although limitations of using a single registration error to calculate LOD are recognized (Fey and Wichmann, 2017), the compromise is used routinely because there are no favorable alternatives (Fan et al, 2015). Fey and Wichmann (2017) already found that greater surface roughness resulted in a greater LOD, while lesser surface roughness had a smaller LOD.…”

“…Scans from different positions need to be transformed into a common external coordinate system from the different internal spherical coordinate systems to obtain a composite point cloud. This is accomplished by scan to scan registration (Fan et al, 2015). A large number of registration methods have been described in the literature, and in general they are distinguished as target based and feature based (Besl and McKay, 1992;Fan et al, 2015;Shen et al, 2016;Liu et al, 2017).…”