Image Recognition-Based Deformation Monitoring Method for Grounding Electrode Foundations

To overcome the problems of long inspection cycles, low accuracy, and the computational and perspective limits of laser modeling without early warning, this paper proposes an image recognition-based monitoring method. A four-layer system-data acquisition, visual reconstruction, dynamic recognition, and monitoring-warning-is established, integrating UAV photogrammetry, SfM-MVS modeling, and CNN/Transformer recognition to achieve precise monitoring and dynamic early warning of fundamental deformations in pole sites. Experiments show significant improvements in precision, response, effectively meeting the deformation monitoring needs of the Shunzhou grounding electrode foundation.