A Study on Real-Time 3D Reconstruction Based on NeWCRFs: A Case Study of Excavation Engineering

Surface surcharge loads around excavation engineering are critical to project stability. However, despite strict regulations, existing monitoring methods lack both precision and real-time capabilities. This study introduces a UAV-based monitoring system that integrates 3D reconstruction and load space extraction to enhance anomaly detection accuracy. The system was tested in ultra-large and ultra-deep excavations in Nanjing's Jiangbei New Area, and a comprehensive surcharge load categories database was established. The NeWCRFs model enables depth estimation from a single RGB image, point cloud generation, and load distribution visualization through spatial projection. Additionally, the system incorporates image enhancement, real-time depth prediction, point cloud modeling, and bird's-eye visualization within a user-friendly interface, facilitating practical deployment. The findings of this study significantly enhance the accuracy of abnormal surcharge load monitoring and provide strong support for excavation safety management.

Keywords: NeWCRFs, 3D reconstruction, Point cloud, Image enhancement, UAV, Excavation engineering.