TLS data georeferencing - error sources and effects

Abstract

Depending on the requirements of a certain engineering task, point coordinates obtained through terrestrial laser scanning (TLS) can be either in a scanner coordinate system (CS) or in the coordinate system of a geodetic control network. When point coordinates in some external CS are needed point cloud georeferencing must be done, i.e. point coordinates have to be transformed from the scanner CS into the desired CS.Different procedures can be followed during the transformation process of point coordinates from one CS to the other and consequently it can be distinguished between several types of georeferencing. The principal classification is into direct and indirect georeferencing and the main difference between the two is that direct georeferencing can (and usually does) give point coordinates in the CS of a geodetic control network instantly in the field, while indirect georeferencing inevitably needs some work to be done in the office in order to obtain these coordinates. Indirect georeferencing is necessarily done in some software and it distinguishes between the process itself being completed in either one or two steps. On the other hand, direct georeferencing does not involve transformation into some intermediate CS whichis the case with the two-step indirect georeferencing. Direct georeferencing essentially mimics the procedure of orienting a total station with respect to a geodetic control network which can be achieved either through backsighting (the “station-orientation” procedure) or resection.This paper briefly presents different georeferencing procedures and related main error sources that cause errors in transformed point coordinates. Additionally, the covariance model for direct georeferencing following the “station-orientation” procedure is verified through statistical analysis of the data collected in the experiment performed in the field. True point position errors calculated as differences between point coordinates obtained from the least squares adjustment of the geodetic control network and those from direct georeferencing of the TLS data are compared with theoretical errors, i.e. model-derived standard deviations of point positions. It is shown that these two setsof errors or, more precisely, the variance of the true errors and the pooled model-derived variance of the control point positions do not feature a significant difference at the confidence level of 99%. This makes us optimistic in terms of possibility of using the reported model for predicting trueerrors of point positions by model-derived standard deviations obtained as a result of direct georeferencing of TLS data following the “station-orientation” procedure.