Generation and Analysis of Urban Morphology in the Qinba Mountains: A GAN-Based Approach for Small Towns in Southern Shaanxi

Abstract

In this study, GAN technology was used to prepare small-town landscape generation and morphological analysis in the Qinba Mountain area. Important for the construction of a standardized dataset, a high-quality remote sensing image and building footprint data vector with adequate diversity and accuracy were utilized. It created exacting, via a GAN model, realistic small town morphologies and gave a sufficiently in-depth quantitative analysis in terms of several morphological indicators of area, perimeter, fractal dimension, longest axis length, circularity, mean distance from centroid to corners, of the convex hull perimeter, and of the aspect ratio. The results showed GAN-generated small town morphologies significantly varying in complexity and diversity. The hierarchical clustering analysis enables to uncover more intrinsic structures of the generated samples and classify the morphological characteristics of different types of towns. Therefore, this research is a verification of the effectiveness of GAN in generating urban morphology, but at the same time, it lays theoretical foundations with proof data for the optimization of GAN models. Therefore, the morphological analysis methods and indicator systems in this paper have a huge reference significance for urban planning and design. In short, this study demonstrated the tremendous potential of GAN technology in the generation and analysis of urban morphology and offered a scientific basis and technical support for small town planning in Qinba Mountain.