Quality Assessment Of Geometric Facade Models Reconstructed From TLS Data

Terrestrial laser scanner (TLS) data plays an important role in 3D modelling of urban scenes. Much research is carried out in the fields of pre-processing, segmentation and geometric reconstruction of laser data of building facades. However, the quality of 3D models obtained from this data is rarely assessed. This paper proposes a qualitative and quantitative assessment approach to evaluate the quality of 3D models of facades constructed with TLS data. Only segmentation and reconstruction operations are considered in this paper, because they cover the main steps in the 3D modelling process. Since every processing step produces an intermediate result, each one must be analysed and compared with an appropriate reference model. A new approach has been conceived to evaluate the precision of the segmentation results. It is based on the calculation of Boolean operators enabling the comparison of planar clusters produced by the segmentation. To evaluate the vector models resulting from the reconstruction step, a method based on quality indices is proposed. It is adapted not only for precision but also for accuracy calculations. The results show that the quality assessment approach is coherent and, moreover, easy to apply.