Urban fabric similarity analysis and classification represent a useful analysis framework in many urban studies. Classical approaches include qualitative descriptions or manually selected indicators, depending largely on the researcher's knowledge. This paper proposes a new method for extracting concise and integrated quantitative indicators to represent urban fabrics via deep learning. Figure-ground images are taken as training data for a convolutional autoencoder (CAE) model, and the outputs of the neurons in the bottleneck layer of the CAE are extracted as the compressed feature vectors (CFVs) to represent the plots. Then, the plots can be compared, clustered and visualized based on these CFVs. In this study, 345 residential plots of Nanjing, China are taken as samples to demonstrate the modeling process, clustering and visualizations. The results show that the CFV is an effective indicator to represent urban fabrics. The generation of the CFV takes into consideration both statistical and geometrical features, with the latter normally described qualitatively as patterns. CFV consists of several aspects of urban fabric without the need to balance the weights. The CFV can serve as a basis for further urban development interpretations, morpho-typology studies and other social, economic and microclimate studies that have relationships with the urban fabric.INDEX TERMS Urban fabric, urban morphology, deep learning, convolutional autoencoder, hierarchical cluster.

No abstract

Although various hierarchical structures have been investigated with respect to the different elements of urban form, the hierarchical spatial order of access from streets to plots and buildings has not been adequately explicated and objectively assessed. In this article, a new method, access structure, is presented to bridge this knowledge gap. Based on Krop’s generic multilevel diagram of urban form, different types of access structure are developed and symbolically represented. They are then quantitatively measured and compared using three metrics and an associated ternary diagram. Subsequently, the new method is tested first in analysing the internal structure of an individual urban block and then in distinguishing urban blocks with different structural characteristics. Eight urban blocks across the city of Nanjing, China, are selected as case examples. The results show that access structure is capable of accurately describing and evaluating complex spatial relationships between streets, plots and buildings. Access structure is potentially a useful method for studying the complex emerging built form of rapidly changing cities, especially in developing countries such as China.

In order to accurately analyse the impact of the rainy environment on the characteristics of highway traffic flow, a short-term traffic flow speed prediction model based on gate recurrent unit (GRU) and adaptive nonlinear inertia weight particle swarm optimization (APSO) was proposed. Firstly, the rainfall and highway traffic flow data were cleaned, and then they are matched according to the spatiotemporal relationship. Secondly, through the method of multivariate analysis of variance, the significance of the impact of potential factors on traffic flow speed was explored. Then, a GRU-based traffic flow speed prediction model in rainy environment is proposed, and the actual road sections under different rainfall scenarios were verified. After that, in view of the problem that the prediction accuracy of the GRU model was low in the continuous rainfall scenario, the APSO algorithm was used to optimize the parameters of the GRU network, and the APSO-GRU prediction model was constructed and verifications under the same road section and rain scene were carried out. The results show that the APSO-GRU model has significantly improved prediction stability than the GRU model and can better extract rainfall features during continuous rainfall, with an average prediction accuracy rate of 96.74%.

No abstract