With the increasing availability of large-scale 3D city models, efficient data storage and transmission formats are essential. While the geospatial community has developed cloud-optimised formats for 2D datasets (binary files that can be efficiently indexed and accessed through HTTP Range requests), 3D city models with complex geometries, attributes, textures, and semantic surfaces still rely on text-based files using the CityGML standard (CityJSON and XML files). In this paper, we present FlatCityBuf, a new compact binary encoding format for 3D city models based on FlatBuffers and CityJSON. Our approach leverages the benefits of FlatBuffers, including cross-platform support, zero-copy data access, and efficient deserialisation, while adhering to the CityGML data model. The addition of spatial and attribute indices enables efficient queries to retrieve partial data. We evaluate the read performance and compression ratios of FlatCityBuf against CityJSONSeq using real-world 3D city models and demonstrate its advantages over existing formats. The results highlight FlatCityBuf’s efficient storage and transfer of 3D city model data, achieving for real-world datasets 10–30% compression compared to the already compact CityJSON format; for deserialisation it is 9–250× faster and uses 2–6× less memory. The schemas and accompanying software for conversion to/from CityJSON are publicly available at <code>https://github.com/cityjson/flatcitybuf under a permissive license</code>.

We introduce a novel deep learning-based framework to interpret 3D urban scenes represented as textured meshes. Based on the observation that object boundaries typically align with the boundaries of planar regions, our framework achieves semantic segmentation in two steps: planarity-sensible over-segmentation followed by semantic classification. The over-segmentation step generates an initial set of mesh segments that capture the planar and non-planar regions of urban scenes. In the subsequent classification step, we construct a graph that encodes the geometric and photometric features of the segments in its nodes and the multi-scale contextual features in its edges. The final semantic segmentation is obtained by classifying the segments using a graph convolutional network. Experiments and comparisons on two semantic urban mesh benchmarks demonstrate that our approach outperforms the state-of-the-art methods in terms of boundary quality, mean IoU (intersection over union), and generalization ability. We also introduce several new metrics for evaluating mesh over-segmentation methods dedicated to semantic segmentation, and our proposed over-segmentation approach outperforms state-of-the-art methods on all metrics. Our source code is available at https://github.com/WeixiaoGao/PSSNet.

Data on the number of floors is required for several applications, for instance, energy demand estimation, population estimation, and flood response plans. Despite this, open data on the number of floors is very rare, even when a 3D city model is available. In practice, it is most often inferred with a geometric method: elevation data is used to estimate the height of a building, which is divided by an assumed storey height and rounded. However, as we demonstrate in this paper with a large dataset of residential buildings, this method is unreliable: <70% of the buildings have a correct estimate. We demonstrate that other attributes and characteristics of buildings can help us better predict the number of floors. We propose several indicators (e.g. construction year, cadastral attributes, building geometry, and neighbourhood census data), and we present a predictive model that was trained with 172,000 buildings in the Netherlands. Our model achieves an accuracy of 94.5% for residential buildings with five floors or less, which is an improvement of about 25% over the geometric approach. Above five floors, our model has only a slight improvement on the geometric approach (5%). The main culprit is the lack of training data for tall buildings, which is uncommon in the Netherlands.

Roads are important for many urban planning applications, such as traffic modelling and delivery vehicle routing. At present, most available datasets represent roads only as centrelines. This is particularily true for OpenStreetMap which provides, among many features, road networks at worldwide coverage. Furthermore, most approaches for creating more detailed networks, such as carriageways or lanes, focus on doing so from sources that are not easy to acquire, such as satellite imagery or LiDAR scans. In this paper we present a methodology to create carriageways based on OpenStreetMap's centrelines and open access areal representations (i.e. polygons) to determine which roads should be represented as two individual carriageways. We applied our methodology in five areas across four different countries with different built environments. We analysed the outcome in a delivery routing problem to evaluate the validity of our results. Our results suggest that this method can be effectively applied to create carriageways anywhere in the world, as long as there is sufficient coverage by OpenStreetMap and an areal representation dataset of roads.

While three-dimensional (3D) building models play an increasingly pivotal role in many real-world applications, obtaining a compact representation of buildings remains an open problem. In this paper, we present a novel framework for reconstructing compact, watertight, polygonal building models from point clouds. Our framework comprises three components: (a) a cell complex is generated via adaptive space partitioning that provides a polyhedral embedding as the candidate set; (b) an implicit field is learned by a deep neural network that facilitates building occupancy estimation; (c) a Markov random field is formulated to extract the outer surface of a building via combinatorial optimization. We evaluate and compare our method with state-of-the-art methods in generic reconstruction, model-based reconstruction, geometry simplification, and primitive assembly. Experiments on both synthetic and real-world point clouds have demonstrated that, with our neural-guided strategy, high-quality building models can be obtained with significant advantages in fidelity, compactness, and computational efficiency. Our method also shows robustness to noise and insufficient measurements, and it can directly generalize from synthetic scans to real-world measurements. The source code of this work is freely available at https://github.com/chenzhaiyu/points2poly.

Recent developments in data acquisition technology allow us to collect 3D texture meshes quickly. Those can help us understand and analyse the urban environment, and as a consequence are useful for several applications like spatial analysis and urban planning. Semantic segmentation of texture meshes through deep learning methods can enhance this understanding, but it requires a lot of labelled data. The contributions of this work are three-fold: (1) a new benchmark dataset of semantic urban meshes, (2) a novel semi-automatic annotation framework, and (3) an annotation tool for 3D meshes. In particular, our dataset covers about 4 km² in Helsinki (Finland), with six classes, and we estimate that we save about 600 h of labelling work using our annotation framework, which includes initial segmentation and interactive refinement. We also compare the performance of several state-of-the-art 3D semantic segmentation methods on the new benchmark dataset. Other researchers can use our results to train their networks: the dataset is publicly available, and the annotation tool is released as open-source.