基于深度学习的水工结构位移亚毫米级监测模型OA

To address the challenges of low image quality,environmental interference,difficult target capture,and insufficient accuracy in machine vision-based displacement monitoring of hydraulic structures,a deep learning-enhanced sub-millimeter displacement monitoring model for hydraulic structures called CC-YOLO is proposed.This model is based on a three-tiered cascaded processing module comprising CARN super-resolution reconstruction,YOLO11 object detection,and DeepSORT object tracking.By enhancing the quality of input images at the sub-pixel level,accurately detecting and localizing targets,and calculating displacement with high precision,it enables non-contact,automated,and precise monitoring of hydraulic structure displacement.The model was applied to the slope displacement monitoring of a hydropower station in Fujian Province.The results showed that compared to other visual monitoring models,this model achieved improvements of 5.95%,5.22%,3.87%,and 3.97%in Precision(P),Recall(R),mAP50,and mAP50:95,respectively.The PSNR and SSIM values of the super-resolution reconstruction results both exceeded 25 and 0.7.The maximum error in displacement estimation was only 0.65 mm,with a root mean square error of just 0.207 mm,and a coefficient of determination(R²)of 0.9702.This model demonstrates high target capture accuracy and displacement estimation precision,offering significant advantages for practical displacement monitoring applications.