Uncertainty in Terrestrial Laser Scanner Surveys of Landslides

Abstract

Abstract:Terrestrial laser scanning (TLS) is a relatively new, versatile, and efficient technology for landslide monitoring. The evaluation of uncertainty of the surveyed data is not trivial because the final accuracy of the point position is unknown. An a priori evaluation of the accuracy of the observed points can be made based on both the footprint size and of the resolution, as well as in terms of effective instantaneous field of view (EIFOV). Such evaluations are surely helpful for a good survey design, but the further operations, such as cloud co-registration, georeferencing and editing, digital elevation model (DEM) creation, and so on, cause uncertainty which is difficult to evaluate. An assessment of the quality of the survey can be made evaluating the goodness of fit between the georeferenced point cloud and the terrain model built using it. In this article, we have considered a typical survey of a landsliding slope. We have presented an a priori quantitative assessment and we eventually analyzed how good the comparison is of the computed point cloud to the actual ground points. We have used the method of cross-validation to eventually suggest the use of a robust parameter for estimating the reliability of the fitting procedure. This statistic can be considered for comparing methods and parameters used to interpolate the DEM. Using kriging allows one to account for the spatial distribution of the data (including the typical anisotropy of the survey of a slope) and to obtain a map of the uncertainties over the height of the grid nodes. This map can be used to compute the estimated error over the DEM-derived quantities, and also represents an "objective" definition of the area of the survey that can be trusted for further use.

Assessment of DEM derived from very high-resolution stereo satellite imagery for geomorphometric analysis

“…The planimetric root mean square error (RMSE (N,E) ) of the map is about 1.2 m and the vertical error RMSE (H) is 0.75 m (Barbarella, Fiani, & Lugli, 2017). The accuracy of these points is clearly lower than the one of the points measured directly on the terrain.…”

Use of DEMs Derived from TLS and HRSI Data for Landslide Feature Recognition

“…Very high return point density was observed in the close vicinity of the scanning stations and in overlapping areas. However, the return point density at the boundaries of the survey area was low to the point that it could no longer be used to develop a reliable terrain model using interpolation techniques [41,42]. The slope causes point density to be higher at the toe of the landslide, nearest to the TLS stations, and lower in the upper part.…”

“…The choice of algorithm parameters can be aided by cross-validation and assessment of the quality of interpolation performed by each set of values to be compared [41]. Using our data sets, we randomly extracted a sample of validation data (5%, used for testing) and fitted the remaining part (95%, used for training) on a grid DEM.…”

“…Then, we computed the differences between the elevation values of each sample point and its corresponding forecast value from to the DEM. We computed a few parameters of the distribution of our testing sample-mean and mode, as well as variance and mean absolute deviation-to evaluate the accuracy of the fitting method [41]. The cross-validation process was run on both data sets.…”

Development of Laser Scanner for Full Cross-Sectional Deformation Monitoring of Underground Gateroads

“…Franz et al [41] used targets to reduce the negative effect of highly vegetation on the scan in landslide mapping. For the purpose of accuracy improvement in mapping landslide with TLS, Barbarella et al [42] evaluated the uncertainty of monitoring data.…”