Influence of fitting models and point density sample in the detection of deformations of structures using terrestrial laser scanning

摘要

This paper addresses the influence of different fitting models and point density samples for the parametric shaping of planar surfaces when used to detect deformations of manmade structures. An experiment was conducted where known angular deformations ranging from 120′′ to 520′′ of arch were applied to a planar structure. Those displacements were measured firstly with an electronic level (Taylor Hobson “Talyvel 5”, precision 0.2′′) and also through normal vectors analysis obtained from the application of different fitting models (total least squares and ordinary least squares) and different size point samples. The results were analyzed from the point of view of the precisions achieved in the fitting, and considering the differences observed between the deformations measured with terrestrial laser scanner and the electronic level. It was shown that the total least squares method performs better than the ordinary least squares method in terms of precision, sensitivity, and efficiency since it detects smaller deformations with higher precision using less quantity of points. Also, it was observed an important improvement of the precision of the measurements when the point density sample grows up to 10 points/dm2, beyond that benchmark, the improvement of the precision is less pronounced. The adoption of an adequate mathematical model to obtain the parameters of representation of a planar feature and a thoughtful definition of the point density sample used, improve the quality of the measurements allowing the detection of sub-millimetric deformations.