LiDAR digital building models for true orthophoto generation

摘要

The importance of digital orthophotos in spatial databases has increased in recent years, since they are an efficient, low-cost and, if properly managed, accurate product. Usually, the generation of orthophotos is carried out using digital terrain models (DTMs); meaning without taking into account vegetation, buildings, and other attached and detached structures. This leads to low accuracies in urban areas, bringing distortions into the image. To avoid this unwanted effect, one must adopt a digital surface model (DSM), as proposed by Amhar et al. (Int Arch Photogrammetry Remote Sens 32(4):16–22, 1998). The method proposed in this paper allows for the creation of true orthophotos by using a DSM to refine the representation of buildings. The pixel size of the DSM must be similar to that of the true orthophoto in order to model the roof edges with sufficient accuracy. This paper presents a new method capable of correcting the roof displacement using an approach based on the integration of several products today available in public administrations, such as a geodatabase, DTMs/DSMs, and light detection and ranging (LiDAR) data. The method is based on a rigorous modelling of simple roofs starting from their 2D projection in the geodatabase, while information about their heights can be obtained using LiDAR data. For some selected simple roofs, automatic modelling can be carried out, in which a robust interpolation method, such as RANSAC, is used to model the pitches identified by a clustering procedure. For complex roofs, where creating a rigorous model in a fully automatic way is not possible, a procedure based on the thickening of a DSM is carried out.