From ruin to restoration: research on historical district facade generation based on shape grammar and stable diffusion

The restoration of historical blocks is a crucial part of heritage conservation and urban renewal. However, traditional restoration methods suffer from insufficient integration of elements and difficulties in data collaboration, which easily lead to results deviating from historical authenticity. This paper proposes a digital restoration framework based on multi-source data fusion. It parses the composition rules of facades using shape grammar, generates facade images conforming to stylistic features with LoRA fine-tuning and ControlNet geometric constraints, and finally realizes 3D texture reconstruction in CityEngine. Taking Heyue Old Street as an example, this method is verified to improve the interpretability of the restoration process and the authenticity of generated outcomes, providing a reference for the dynamic conservation and renewal of historical blocks.