Defect Detection of Transmission Lines Based on Fusion of Thermal Imaging Information

In light of the susceptibility of abnormal heating in infrared images to disturbances from factors such as time and background, a semantic enhancement method for infrared images based on optical information was proposed. Firstly, the semantic information of the transmission line is extracted from the optical image using the semantic segmentation algorithm of U-net. Subsequently, the intrinsic reference matrix and distortion coefficient of the optical wide-angle camera are calculated through internal parameter calibration. The optical image and semantic features are then matched with the infrared image using affine transformation. Finally, the Accelerated Robust Feature Algorithm (SURF) is employed to match the pixels of the photothermal image and enhance it based on the semantic features. Finally, the YOLO v8 network is used to verify the information fusion effect, and the results show that the proposed semantic enhancement algorithm improves the mAP50 by 13.9%.