Research on Convolutional Neural Network Method for Transmission Line Abnormal Sound Identification

At present, most of the hidden troubles in transmission lines are detected by image and video technology, however, there are too many monitoring targets, due to the influence of weather changes and lighting, and other factors, it is difficult to accurately monitor the hidden troubles of transmission lines only by video images, this paper presents a convolutional neural network method to identify the abnormal sound in transmission lines. The audio data collected by the audio-visual acquisition device is processed by the convolutional neural network to identify the potential audio data, filter the non-potential audio data, and improve the speed of potential danger analysis. Through the training and testing of sound multi-label classification and detection algorithm, the validity of the method is verified.