Infrastructure design is conducted in a 3D Cartesian coordinate system with the assumption that the Earth is flat and that the scale is constant over the entire project area. Map projections are commonly used to georeference the designed geometries before constructing them on the surface of the Earth. The scale in a map projection varies depending on the position in the map plane, which leads to scale distortions between the designed geometries and the geometries staked out for construction. These distortions are exaggerated for large longitudinal projects such as the construction of roads and railroads because the construction site spans a larger area. Building Information Modeling (BIM) is increasing in popularity as a way to manage information within a construction project. Its use is more widespread in the building industry, but it is currently being adopted by the infrastructure industry as well. The open BIM standard IFC (Industry Foundation Classes) has recently developed support for alignment geometries, and full support for disciplines such as road and railroad construction is underway. This study tests whether the current IFC standard can facilitate georeferencing with sufficiently low distortion for the construction of infrastructure. This is done by performing georeferencing using three different methods, all using the information provided in the IFC schema, and by calculating the scale distortions caused by the different methods. It is concluded that the geographic capabilities of the IFC schema could be improved by adding a separate scale factor for the horizontal plane and support for object-specific map projections.

Capturing geographic information from a mobile platform, a method known as mobile mapping, is today one of the best methods for rapid and safe data acquisition along roads and railroads. The digitalization of society and the use of information technology in the construction industry is increasing the need for structured geometric and semantic information about the built environment. This puts an emphasis on automatic object identification in data such as point clouds. Most point clouds are accompanied by RGB images, and a recent literature review showed that these are possibly underutilized for object identification. This article presents a method (image-based point cloud segmentations - IBPCS) where semantic segmentation of images is used to filter point clouds, which drastically reduces the number of points that have to be considered in object identification and allows simpler algorithms to be used. An example implementation where IBPCS is used to identify roadside game fences along a country road is provided, and the accuracy and efficiency of the method is compared to the performance of PointNet, which is a neural network designed for end-to-end point cloud classification and segmentation. The results show that our implementation of IBPCS outperforms PointNet for the given task. The strengths of IBPCS are the ability to filter point clouds based on visual appearance and that it efficiently can process large data sets. This makes the method a suitable candidate for object identification along rural roads and railroads, where the objects of interest are scattered over long distances.

The adoption of building information modeling (BIM) in the architecture, engineering, and construction (AEC) industry is increasing the need for accurate object-oriented models of existing assets in the built environment. Moreover, 3D mesh models of man-made objects used in the design process are becoming more easily available. This paper explores synergies between the design and parsing of the built environment by procedurally generatingsynthetic point clouds from 3D mesh models. The synthetic point clouds were used as training data for PointNet,a neural network for end-to-end point cloud classification and segmentation, and the trained network was validatedusing "real" point clouds created through laser scanning. The conclusions from the work are that there is potential for the tested approach and that there are areas with clear needs for future research. All code and point cloudsamples used to generate training data are publicly available.