A topography-aware approach to the automatic generation of urban road networks

Abstract

Existing deep-learning tools for road network generation have limited applications in flat urban areas due to their overreliance on the geometric and spatial configurations of street networks and inadequate considerations of topographic information. This paper proposes a new method of street network generation based on a generative adversarial network by designing a pre-positioned geo-extractor module and a geo-merging bypath. The two improvements employ the complementary use of geometric configurations and topographic features to automate street network generation in both flat and hilly urban areas. Our experiments demonstrate that the improved model yields a more realistic prediction of street configurations than conventional image inpainting techniques. The model’s effectiveness is further enhanced when generating streets in hilly areas. Furthermore, the geo-extractor module provides insights from the computer vision perspective in recognizing when topographic information should be considered and which topographic information should receive more attention.

Keywords:

• Street network generation
• topographic information
• computer vision
• generative adversarial network
• planning support systems

Acknowledgments

The authors are also grateful for the anonymous reviewers and editor for their constructive comments, which helped improve the quality of the paper.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data and codes availability statement

The data and codes that support the findings of this study are available at the following link: https://doi.org/10.6084/m9.figshare.19606960.v1.

Notes

1 Note that the impacts of planning guidance have been explored in Fang et al (2021) and therefore is out of the scope of this research.

2 More test results can be found in Supplemental Material D.

3 Specifically, we cropped one data sample from the multi-channel base maps of Rome and Siena every 25 pix in both horizontal and vertical direction, forming a new dataset with 6,800 samples for GE analysis.

Additional information

Funding

Zhou Fang appreciates the support from the China Scholarship Council and the Cambridge Commonwealth, European & International Trust through a CSC–Cambridge Trust scholarship (CSC No. 201808060082), Magdalene College via Ng Fourth-Year PhD Bursary, and the Alibaba Group through the Alibaba Research Intern Program. Ying Jin wishes to acknowledge funding support from the Tsinghua University Initiative Scientific Research Program via the Tsinghua–Cambridge research collaboration. Tianren Yang wishes to thank the support from the HKU-100 Scholar Fund.

Notes on contributors

Zhou Fang

Zhou Fang received his B.Eng. and M.Phil. degrees from the University of Hong Kong and the University of Cambridge, respectively. He is currently pursuing his Ph.D. degree at the Martin Centre for Architectural and Urban Studies, Department of Architecture, the University of Cambridge. His current research interest lies in deep learning and its applications in the fields of geographical information science and urban studies.

Jiaxin Qi

Jiaxin Qi received his B.Eng. degree from the University of Science and Technology of China in 2018. He is currently pursuing a Ph.D. degree at the School of Computer Science and Engineering, Nanyang Technological University. He is interested in the theory of deep learning and computer vision.

Lubin Fan

Lubin Fan is a Senior Algorithm Engineer at Alibaba Damo Academy with research interests including urban reconstruction, procedural modelling, image and video processing, and geometric processing. He received his B.Sc. and Ph.D. degrees in Mathematics from Wuhan University in 2009 and Zhejiang University in 2014, respectively. From 2014 to 2017, he was a postdoctoral fellow at the Visual Computing Center, King Abdullah University of Science and Technology (KAUST).

Jianqiang Huang

Jianqiang Huang is currently the Director of the Alibaba DAMO Academy. His research interests focus on the visual intelligence in the city brain project of the Alibaba Group. He was a recipient of the Second Prize of the National Science and Technology Progress Award in 2010. and owned the title of ‘Outstanding Scientific and Technological Worker’ awarded by the Chinese Institute of Electronics in 2021.

Ying Jin

Ying Jin is a Professor of Architecture and Urbanism at the Department of Architecture at the University of Cambridge. He is the Leader of the Cities and Infrastructure Research Group and the lead convenor of the international symposium on applied urban modelling since 2011.

Tianren Yang

Tianren Yang is an Assistant Professor in the Department of Urban Planning and Design at the University of Hong Kong. He is interested in developing advanced urban analytics and modelling to decipher how city regions evolve in terms of land development, human activity, and social welfare. He earned his Ph.D. in applied urban modelling from the University of Cambridge and is a chartered urban planner in the UK and China.