학술논문
Transmission Line Image Object Detection Method Considering Fine-Grained Contexts
Document Type
Conference
Author
Source
2020 IEEE 4th Information Technology, Networking, Electronic and Automation Control Conference (ITNEC) Automation Control Conference (ITNEC), 2020 IEEE 4th Information Technology, Networking, Electronic and. 1:499-502 Jun, 2020
Subject
Language
Abstract
It takes a huge amount of works to take pictures of transmission line towers and check electrical fittings manually. In spite of the introduction of deep learning technology to transmission line inspection, it is not well utilized that fine-grained contexts on components in state-of-the-art research. On the basis of region-based fully convolutional network (R-FCN), a novel object detection method is proposed considering fine-grained contexts among electrical fittings. Deformable convolution layers and squeeze-and-excitation (SE) blocks are adopted in the detection method. A comparison experiment is conducted on a transmission line aerial inspection dataset. The proposed method shows better accuracy than R-FCN.