Dual-path framework analysis of crack detection algorithm and scenario simulation on Fujian Tulou surface

Fujian Tulou, a UNESCO World Heritage Site, is highly vulnerable to environmental and anthropogenic stresses, with its earthen walls prone to surface cracking that threatens both structural stability and cultural value. Traditional manual inspection is inefficient, subjective, and may disturb fragile surfaces, highlighting the need for non-destructive and automated solutions. This study proposes a dual-path framework that integrates lightweight crack detection with independent physical simulation. On the detection side, a YOLOv12-MLE model incorporating a MobileNetV4 backbone, improved multi-head cross-attention, and an enhanced EIoU loss function was developed, achieving high accuracy while remaining lightweight and deployable on edge devices. On the simulation side, a two-layer RFPA3D model was employed to parameterize loading ratio (λ) and overlay thickness (t), reproducing the four-stage crack evolution process of initiation, propagation, coalescence, and saturation. Results demonstrate that crack morphology is governed primarily by λ, while crack density is influenced by t, consistent with observed patterns on Tulou facades. Quantitative validation on paired samples showed that the proposed framework outperformed the baseline in skeleton overlap, spacing accuracy, and orientation consistency. Overall, this dual-path strategy provides an effective and feasible approach for non-contact, standardized crack documentation and mechanistic interpretation, supporting preventive conservation and future data–physics integration.