Open AccessTransmission Line Scene Classification Based on Light-VGGNet
Author(s) -
Hongxing Wang,
Zheng Huang,
Bin Liu,
Xiang Huang,
Wei Han,
Hongchen Li
Publication year - 2020
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1631/1/012038
Subject(s) - computer science , artificial intelligence , transmission (telecommunications) , computer vision , line (geometry) , telecommunications , mathematics , geometry
In recent years, power departments have gradually utilized UAV to carry out regular inspections of transmission lines to ensure the safety and stability of power systems. However, a large number of images without useful information are usually captured during the UAV inspection. These images that contain no transmission line information are transmitted to the ground station with informative images, which leads to the surge of workload. To solve this problem, we propose an intelligent image filtering method based on the VGGNet, which is called Light-VGGNet. Firstly, this paper collects the aerial images captured by cameras during the UAV inspection and builds an aerial dataset of transmission line scenes. Then, the Light-VGGNet is proposed, which achieves much lower memory consumption and faster running speed than those of the VGGNet. We use the aerial image dataset to train the proposed network. Finally, the best weight is loaded and utilized to predict the label of the test dataset.