Terrestrial laser scanners and contrast targets: Effect of algorithm, resolution, distance, and target rotational orientation on center coordinate

Terrestrial laser scanners (TLSs) are portable dimensional measurement instruments that produce a 3D point cloud by measuring the distance and two angles to points in the scene. Contrast targets are sometimes used to evaluate the performance of these scanners by comparing inter-target distances to measurements made using higher accuracy instruments or to register scans from different positions of the TLS. The choice of algorithm, distance of the target from the TLS, scan resolution, and rotational orientation of the target contribute to errors in the target center coordinates. We explored this problem using an artifact consisting of spheres and a contrast target that allowed us to monitor the center of the contrast target in a coordinate system established using the spheres, where the assumption is that the sphere centers are unaffected by factors mentioned earlier and therefore provide a stable coordinate system. We considered three different TLSs and four algorithms to detect contrast target center (three custom developed and one from the manufacturer). For each TLS, we performed scans at two distances, two resolutions, and three rotational orientations of the target, for a total of 12 cases (with 10 repeated scans per case, total of 120 scans). We calculated two response variables for each case – inter-case variation d and intra-case variation s. For TLS A, low resolution scans produced the worst results (large d and s values) for the OEM algorithm, with d and s as large as 2 mm and 0.3 mm, respectively, for one of the cases. For TLS B, target orientation of 90° produced the worst results for the OEM algorithm, with d and s as large as 2 mm and 1.3 mm, respectively, for certain cases. For TLS C, only the far distance, low resolution, and 90° target orientation case produced the worst result for the OEM algorithm, with d and s values as large as 0.3 mm and 0.6 mm, respectively. Our custom algorithms clearly outperformed the OEM algorithms for TLS A and B and while the OEM algorithm marginally outperformed the custom algorithms for TLS C.