Super-resolution reconstruction method of transmission line key components image based on SRGAN

Haixia Ma,Zhongxing Li,Na Shen,Jiaqi Huang
DOI: https://doi.org/10.1117/12.2611359
2021-10-14
Abstract:Aiming at the problem of unmanned aerial vehicle inspection images being susceptible to environmental interference during shooting, resulting in blurry image capture and inability to accurately identify defects in key components of transmission lines, this paper uses SRGAN to super-resolution reconstruction of low-resolution inspection images to improve image quality to meet the needs of deep learning algorithms or manual accurate recognition of line defects. First, a high-resolution image data set of key components of the transmission line is produced, and the data set is obscured as a low-resolution image data set. Then the PaddlePaddle framework is used to build the SRGAN super-resolution network model to perform super-resolution reconstruction on the low-resolution data. In model training, the model parameters are optimized according to the training situation, the optimal model is obtained, and the reconstruction experiment on low-resolution images is performed. The experimental results show that the image generated by SRGAN is similar to the high-resolution image in sharpness, and has achieved good results.
What problem does this paper attempt to address?