Transmission Lines Monitoring Based on Convolution Neural Network and Edge Computation

Transmission lines ordinary inspection is an important work for power companies to ensure its safety and stability. With the development of ubiquitous power Internet of Things, realizing intelligent analysis of transmission lines by edge computing is the development direction of smart grid. Convolutional neural networks combining with an edge computation platform are used for feature extraction in this paper. The network structure comprehensively considers some features, such as complex image scenes, the difficulty in extracting hidden dangers type features, and large target size variation spans. The proposed model uses multi-scale prediction and feature fusion technology to detect common hidden dangers in the transmission lines surroundings. Experiments show that the mean average precision (mAP) of our model is 89.1%, and the detection speed is about 1.25 frames per second (FPS). The higher detection accuracy lays a solid foundation for the image intelligent processing algorithm of transmission lines and the edge computation platform.