Calibration of a Terrestrial Laser Scanner - Institute of Geodesy and ...

6.2 Static Application: Rock Engineering Applications 117Each free positioning of the total station was determined by surveying all four reference points, cf. Figure6.9. The calculation of 3D coordinates of the object points was carried out in an adjustment.The achievedaccuracy was less than 0.3 mm. The first measurement of the object points was used as an initial or referencemeasurement (session 0). Displacementswere calculated by subtractingthat occurred between the initial and the «th measurement sessionthe coordinates of session i from the initial coordinates of session 0. Theresulting accuracy of the derived displacements of the object points in each coordinate direction (x, y, z)was quantified with less than 0.5 mm. Therefore, the accuracy of 3D displacements for each pointi, with respect to session 0, were specified with less than 1 mm.of sessionThe object points can also be defined by spheres, instead of prisms, attached on the bolts installed, cf.Figure 6.7.Spheres are well-adapted for laser scanning because of their attractive properties regardingvisibility and deriving center points, cf. Section 3.1.5. Each sphere was scanned and, based on the resultingpoint cloud, the coordinates of the center point of the sphereswere calculated. Thisapproachresults inanaccuracy of the center point of less than 3 mm, cf. Section 3.6.2. In addition, the center points have tobe referred to the local reference frame. Nevertheless, it is not always possible in practicereference points and the object points from the same scanner position.to scan both theFirst, the distances from the laserscanner to the reference points are usually too long to achieve the required accuracy. Second, the referencepoints are not always visible from the position of the laser scanner. Looking at Figure 6.9, it can be seen thatthese problems occurred during the surveying work in the laboratory, which required positioning the laserscanner inside the niche. To solve this problem, intermediate reference points were set up temporarily andclose to the laser scanner using tripods. These reference points were also surveyed bythe total station andwere included in the local reference frame. The accuracy of the position of the laser scanner can be specifiedwithin 3 mm. The resulting accuracy of the derived displacements5 mm. Consequently, the accuracy of 3D displacements for each object pointin one coordinate direction was less thanis within 9 mm.Understanding the actual rock mass behaviour can be greatly improved if the distribution of the displace¬ments along the excavation surface is investigated instead of considering the displacementsof a few dis¬crete points. For that purpose, point clouds representing surfaces scanned at different phases of the excava¬tion can be used. Several software packages allow for the 3D comparisons of pointclouds and surfaces. Inthis case study, the software Geomagic by Raindrop Geomagic Inc. was used. The generation of time-lapsedisplacement maps requires some processing beforehand.First, points representing blunders have to bedetected and deleted automatically or manually. Then, the noise is reduced by means of a filtering process.This is an essential step as the noise due to the natural limits of scanning affects greatly the qualitypoint cloud, by making sharp edges dull and making smooth surfaces rough.form arrangement of points. Subsequently, the processingof theThe result is a more uni¬entails the conversion of the initial or referencepoint cloud into a surface model that consists of small triangles (TIN, Triangular Irregular Network). Thissurface model represents the reference object that can be further processed, if required, e.g. by deletingnon-contiguous intersecting triangles, filling holes or surface smoothing. Finally,the residuals of a test ob¬ject described by a point cloud or a TIN representing the same region at a different time can be computedby comparing this test object with respect to the reference object. This operation is possible onlyobject has been transformed previously into the same reference system as the reference object.if the testAs another possibility, the use of the least squares 3D surface and curve matching algorithm, developedby [Grun and Akca, 2005], may help the interpretation and detection of surface displacements based on3D point clouds.Therefore, the possible displacements of objects, described by 3D point clouds, can becharacterized by a 3D translation vector and a 3D rotation vector. Thus, new insightsrock masses can be obtained.into the behaviour of