The geometry of terrestrial laser scanning: identification of errors, modeling and mitigation of scanning geometry

Over the past few decades, Terrestrial Laser Scanners are increasingly being used in a broad spectrum of applications, from surveying to civil engineering, medical modeling and forensics. Especially surveying applications require on one hand a quickly obtainable, high resolution point cloud but also need observations with a well described quality, from which it is possible to reliably derive the quality of the end-product. As any measurement, TLS scans are subject to measurement noise. Currently, the manufacturers provide documentation containing only global technical specifications including precision of measurements performed on reference surfaces under laboratory conditions. After brief introduction of the principal of Laser Scanning, in this thesis an overview of the major quality influencing factors is provided, grouped in four main categories: (i.) scanner mechanism, (ii.) atmospheric conditions and environment, (iii.) object properties and (iv.) scanning geometry. In many cases, the user has limited control on the scanner mechanism, the atmospheric conditions or the object properties. The only factor on which the user has control on is the scanning geometry, as the user determines the scan location and thereby the view-point of a point cloud. This dissertation presents the research on the influence of scanning geometry on the point cloud quality. This thesis proposes a theoretical study of the scanning geometry effects on individual point quality, as well as practical assessments. The impact of scanning geometry on individual point quality is analyzed, based on local planar features. The quality investigated in this thesis relates to the random errors or precision of individual points and does not deal with systematic errors or biases. Different planar fitting techniques are presented and compared. The quality of each local fit is described using a Least Squares estimation. The main quality describers used in this work are presented for each method. By using these quality describers, the influence of the scanning geometry on the point quality is characterized both quantitatively and qualitatively. The scanning geometry is defined using two parameters: the incidence angle and the range. The incidence angle is defined as the angle between one laser beam vector and the normal vector to the surface. The range is defined as the distance between the scanner and the surface. It is shown that and how the received signal strength of the measurements decreases with increasing incidence angle and range. The presented approach allows the quantification of the contribution of noise induced by the scanning geometry, based solely on point cloud data. No additional or external measurements are needed. The contribution of the two scanning geometry parameters on the point quality has been quantified using contribution coefficients. The effect of scanning geometry on the point quality is quantified and tested on a reference test board and two point clouds sampling a standard room. It is shown that the theoretical models developed are consistent with this experimental assessment. It is shown that it is possible to reduce the total error of the measurements by placing the scanner at another position in the room, which is not necessarily an obvious position. Inspired by these results, a new method that determines near optimal view-points in a scene based on terrestrial laser scanner capabilities is presented. Using a simple approach, an improvement of the measurement set-up can be easily achieved using a small amount of computation, memory and time