Optimization of Terrestrial Laserscanning (TLS) Networks based on Sensitivity Analysis

点击翻译按钮获取中文摘要

Sensitivity analysis as understood in engineering geodesy usually deals with the ability to detect small deformations based on an assumed or expected deformation model (rigid body movements and distortions). In this project, deformations will be monitored by Terrestrial Laser Scanning (TLS) from different scanning sites (viewpoints). The challenge is to develop a multi-criteria optimization method for a first and second order geodetic design of a TLS network. This covers the number and positions of the viewpoints as well as the scanning settings (e.g. scan rotation velocity, angular resolution). The optimization should cover completeness and sensitivity. All this information can be provided in the planning phase for the viewpoint network. The developed optimization software should work for any given object geometry and any expected deformation. Assuring completeness means that all surfaces, edges and corners of the object are visible in the registered joint point cloud and sufficient overlap among the individual scans for registration is guaranteed. The completeness optimization using Greedy algorithms, mixed linear programming or metaheuristic approaches deliver a minimum number of TLS viewpoints together with their positions. This represents the initial solution for the optimization with respect to sensitivity. The objective function for the final optimization is the determinant of the covariance matrix in the direction of the deformation model. Therefore, the uncertainty of the object points acquired from each scanning site plays a substantial role. Here, scanning parameters like scan rotation rate and angular resolution, object-scanner distance and incidence angle, are considered. These parameters influence the sensitivity with respect to deformation models. A metaheuristic approach (e.g. genetic algorithms, simulated annealing) will be used for the sensitivity optimization. During the optimization procedure, the number of viewpoints and the spatial density of their possible positions are varied until the global minimum is reached. Since two optimization objectives may lead to a conflict, a number of possible (Pareto-optimal) solutions have to be provided. This target conflict may be further extended in case of additional objective functions (multi--dimensional case); e.g. by minimum time for scanning the network, thus a Pareto front of possible optimal-solutions is determined. Additionally, radar plots are created for decision support. In contrast to the research up to now, this project simultaneously aims at optimization with respect to minimal detectable 3D deformations and 3D completeness. Hence, a multi-criterial optimization will be investigated and a viewpoint planning tool will be developed.