Multi-scale transmission line defect identification for new power system based on convolutional neural network

In the face of multiple multi-scale transmission line defects, existing methods have problems with low identification accuracy and long identification time. Therefore, this paper proposes a multi-scale transmission line defect identification for new power system based on convolutional neural network. Firstly, use unmanned aerial vehicles to collect multi-scale images of transmission line defects and correct the collected images. Secondly, this article establishes a unified data centre for storing, managing, and analysing this data. Then, convolutional neural networks are used to extract defect features from multi-scale transmission line images. Finally, the extracted defect features are input into the support vector to complete the identification of multi-scale transmission line defects. The experimental results show that the proposed method can accurately identify various types of transmission line defects and shorten the identification time.