Transmission Line Information Extraction from Images Collected by UAV based on Generative Adversarial Networks

Based on PyTorch development platform, this paper builds the Generative Adversarial Networks (GAN) model. Through the preprocessing, label making, network training and algorithm improvement of UAV aerial images, this paper completes the deep-learning of transmission line feature information, solidifies the Generation Network parameters, and realizes the goal of automatic extraction of transmission line information from UAV images. Based on the Deep Convolution Neural Network, a multi generator GAN model is proposed. The cooperative working mechanism is introduced between the generation networks to speed up the model to obtain information and reduce the amount of parameters. The Wasserstein distance is introduced into the loss function of the model to avoid the problems of gradient disappearance and training instability in the process of GAN training. Through experimental analysis, it is proved that this method has a good reference for extracting transmission line information from high-resolution UAV images.