|M.Sc Student||Shlomo Torgman|
|Subject||Accuracy of Laser Scanning Systems and the Components|
|Department||Department of Civil and Environmental Engineering||Supervisor||Professor Filin Sagi|
|Full Thesis text - in Hebrew|
The emergence of terrestrial laser scanning as a tool for dense and accurate 3D mapping has seen a growing number of applications, which utilize this technology for a diverse set of purposes. By directly emitting energy pulses at high frequency and measuring the travel time to the object surface, laser scanners can gather a large amount of three-dimensional data rapidly.
Despite the potential and the high level of detail, laser scanner accuracy is usually described by manufacturers’ specifications with limited detail of the actual affecting factors. Therefore, the importance of studying the accuracy of laser scanners and understanding the factors that influence it cannot be understated.
In order to examine their accuracy and reliability, this thesis develops methods and procedures for analyzing laser scanners accuracy and understanding factors that affect them. The study consists of two parts: the first concerns examination of the factors influencing the laser scanner measurement accuracy under different scenarios; the second concerns construction of calibration procedures for extraction of systematic errors within terrestrial laser scanners.
Experiments performed for assessing the accuracy estimates consisted of indoor and outdoor scans and analyses of the vertical accuracy along a variety of ranges. These results, which were inferred for different scanners show high confidence in the height measurements almost irrespective of the measured range. Assessment of the vertical accuracy was performed by evaluation of common targets measured at different inclination angles. These analyses have demonstrated high accuracy level as well. Evaluating the scanners performance for different albedo, aspects, incidence angles, and different geometry, was then performed by testing planar as well as cylindrical objects positioned in different settings. For mapping projects evaluation, polygonal mapping related experiments were performed. They have shown how local, rather than global, placement of the registration targets can decrease the registration accuracy, without control on the propagated errors. Placement of multi-scan related targets managed to limited the error and provide reliable solutions. In addition, different settings of external control were evaluated.
This thesis also developed a scanner calibration model. Three different settings were evaluated, differing from one another by their level of complexity. The assumed fixed position of the scanner and known coordinates of the control points. The second relaxed this model by estimating also the scanner orientation as well as the calibration parameters. Finally, the third model considered a configuration in which targets are scanned from at least two unknown positions and the reference coordinates of the targets is unknown as well. Clearly, the first configuration yielded the better parameter estimation among the three. By proposing a relative transformation, a relatively compact model could be derived for the third configuration in which only the calibration parameters and the relative orientation parameters were estimated. Experiments with this model shown that while quality of the estimation was lower than the first two, reliable calibration parameters could still be derived. Application of the model in a real-world experiment managed to estimate those parameters with high accuracy.