控制与信息技术Issue(4)：67-73,7.DOI:10.13889/j.issn.2096-5427.2025.04.010
基于域迁移学习的接触网鸟巢检测研究
Study of Bird Nest Detection on Catenaries Based on Domain Transfer Few-Shot Learning
孙元波1
作者信息
- 1. 国能朔黄铁路发展有限责任公司,河北 沧州 062350
- 折叠
摘要
Abstract
In the operation of electrified railways,the potential attachment of bird nests on catenaries poses a safety hazard.Inspecting catenaries for bird nests before train departure is considered essential to ensure the stable and safe operation of trains.Given that manual inspection is not only inefficient,but also consumes a lot of manpower and material resources,intelligent inspection has become the current trend.However,due to the sporadic nature of building bird nests on catenaries,it is quite challenging to follow conventional intelligent detection methods,which requires extensive collection and tagging of bird nest data in numerous catenary scenarios for subsequent learning.This paper focuses on developing a detection model for bird nests on catenaries based on an asymmetric adaptation paradigm for few-shot domain adaptive object detection(AsyFOD).This model integrates deep learning technology to enable identification of bird nests,achieving accurate and efficient detection even with limited tagged data.Experimental results show that the model can achieve a detection rate of 93.7%and a false positive rate of 3.1%,which provides a more feasible solution for practical application scenarios.关键词
鸟巢检测/深度学习/域迁移/小样本学习/AsyFODKey words
bird nest detection/deep learning/domain migration/few-shot learning/AsyFOD分类
交通工程引用本文复制引用
孙元波..基于域迁移学习的接触网鸟巢检测研究[J].控制与信息技术,2025,(4):67-73,7.
