Design of conditional control generation based on regional feature quantification: practical investigation of diffusion models in developed urban areas

Shencheng Xu^*Haitao Jiang

• Shandong Jianzhu University, Jinan, China

Against the backdrop of rapid urbanization, the homogenization of urban form has emerged as a critical challenge, which erodes the cultural identities and functional diversity of cities. Existing workflows for urban form design predominantly rely on subjective experience, qualitative analysis, and quantitative evaluation. However, these approaches have failed to strike a balance between the representation of urban morphological attributes and the preservation of spatial diversity. To address these limitations, this study proposes a deep learning-based conditional generative control model, with metropolitan areas of developed cities worldwide as the research context. First, we established a quantitative evaluation framework for urban form based on landscape pattern indices, which objectively characterizes urban form across five dimensions: shape, scale, compactness, fragmentation, and proximity. Second, we constructed a high-quality urban morphology dataset and verified the correlation between landscape pattern indices and the features of generated urban forms. Finally, we evaluated the accuracy and scalability of the proposed model. The results show that the Percentage of Like Adjacencies (PLADJ) index exhibits the strongest correlation with the model-generated urban morphological features in over 90% of sample images, while the model achieves an average accuracy of 85.72% in generating urban morphological indicators across the five dimensions. This study reveals that although the definition of single-dimensional indicators is context-dependent, they are closely correlated with the overall characteristics of urban form. This research makes two key contributions: Methodologically, it innovatively integrates landscape pattern quantification with conditional generative models, breaking the constraints of traditional subjective design and fragmented quantitative analysis. Practically, it provides a reliable technical tool for quantitative urban morphology research, which can support decision-making in urban planning and design practices.