Insulator Detection Based on Deep Learning Method in Aerial Images for Power Line Patrol
Zheng Huang, Hongxing Wang, Bin Liu, Jie Zhu, Wei Hong Han, Zhaolong Zhang · 2021 11th International Conference on Power and Energy Systems (ICPES) · 2021
Insulators play an important role of supporting wires and insulating in power lines, and insulator detection using aerial images has became a trend in power line patrol. However, traditional visual detection methods are usually less accurate and less robust due to the complex image background. To solve this problem, this paper propose a novel and efficient deep learning based method to detect insulator in aerial image. First, the ResNet network pre-trained on ImageNet dataset is adopted as the backbone network of the proposed network. Then, a cascaded convolution module is designed to extract composite insulator features corresponding to multiple receptive fields. To make the proposed network to learn the most powerful insulator features, SPP module and SE module are introduced into the detection branches of the proposed network. Experimental results verify that the proposed network achieves higher accuracy than five existing networks.