Generating 3D city models without elevation data

Abstract

Elevation datasets (e.g. point clouds) are an essential but often unavailable ingredient for the construction of 3D city models. We investigate in this paper to what extent can 3D city models be generated solely from 2D data without elevation measurements. We show that it is possible to predict the height of buildings from 2D data (their footprints and attributes available in volunteered geoinformation and cadastre), and then extrude their footprints to obtain 3D models suitable for a multitude of applications. The predictions have been carried out with machine learning techniques (random forests) using 10 different attributes and their combinations, which mirror different scenarios of completeness of real-world data. Some of the scenarios resulted in surprisingly good performance (given the circumstances): we have achieved a mean absolute error of 0.8 m in the inferred heights, which satisfies the accuracy recommendations of CityGML for LOD1 models and the needs of several GIS analyses. We show that our method can be used in practice to generate 3D city models where there are no elevation data, and to supplement existing datasets with 3D models of newly constructed buildings to facilitate rapid update and maintenance of data.

Parcel Delivery in an Urban Environment Using Unmanned Aerial Systems: A Vision Paper

“…Open standards like Geographic Markup Language (GML) and its specialized version CityGML offer the means to develop and distribute such 3D city models (Belussi et al, 2015;Gröger and Plümer, 2012). However, there are still open research issues including the utilization of accurate elevation data (Biljecki et al, 2017) and how these models could effectively be incorporated in an UAS.…”

Overview of the Croatian Land Administration System and the Possibilities for Its Upgrade to 3D by Existing Data

“…Section 5 provides an overview of how certain 3D situations are represented on 2D cadastral maps with topographical signs. These situations could be combined with 3D data generated without elevation data as explained by Biljecki et al [20], where it is simple to generate a 3D model from a cadastral map and the number of storeys. The height of the building was estimated by multipling the number of storeys with the average height of a single storey.…”

“…task. Initially, 3D models could be generated from the existing sources previously mentioned, or following the approach described by Biljecki et al [20], or through the semi-automatic approach explained by Pouliot et al [22]. Since existing 3D data are heterogenuous, a machine learning paradigm could be used to overcome slight differences in the data models across various sources.…”

“…For such buildings, a 3D model (LoD 1) could be created as explained in Reference [20]. Buildings and building parts that have technical documentation (for example, new buildings which have valid construction and usage permits) will be represented in 3D profile (LoD 2), while building units will be represented in 2.5D (footprint and height).…”

“…Mentioned studies have dealt with the integration of cadastral data with classical land survey data such as tachymetry or GNSS, but none of them address the modelling of 3D data from Light Detection and Ranging (LIDAR) or other sensors that produce 3D data. If 3D models are generated without elevation data (as explained in Biljecki et al [20]) or from existing data gained from legalization (such as topographical databases, etc. ), these sources and their relationship to spatial units should be modelled via the LADM LA_SpatialSource class.…”