一种高分辨率遥感影像高速铁路及伴行公路智能识别方法  

An Intelligent Recognition Method of High-speed Railways and Accompanying Highways in High Resolution Remote Sensing Images

作　　者：卞磊[1] 黄雪亭 姚家起 BIAN Lei;HUANG Xueting;YAO Jiaqi(Jinan Institute of Surveying and Mapping,Jinan 250000,China;Jinan Qilu Construction Project Management Co.,Ltd.,Jinan 250000,China)

机构地区：[1]济南市勘察测绘研究院,山东济南250000 [2]济南齐鲁建设项目管理有限责任公司,山东济南250000

出　　处：《测绘与空间地理信息》2025年第2期143-145,149,共4页Geomatics & Spatial Information Technology

摘　　要：针对现有技术提取高速铁路及其伴行公路智能化水平低的问题,本文提出一种基于深度学习U^(2)-Net网络模型的高速铁路及其伴行公路智能提取方法,充分挖掘光学遥感影像几何与纹理信息,精细化地提取高速铁路及伴行公路。实验选用成渝线高分2号遥感影像卫星数据,研究结果表明:U^(2)-Net网络模型能够精准地识别与提取高速铁路及其伴行公路,交并比(Intersection Over Union,IoU)达到了73.48%,精度为90.2%,充分说明U^(2)-Net智能识别高速铁路及伴行公路的可行性与可靠性。相关研究成果可为城市规划发展及铁路减灾防灾提供一定的技术参考。In view of the low level of intelligence in extracting high-speed railways and their accompanying highways by existing technologies,this paper proposes an intelligent extraction method for high-speed railways and their accompanying highways based on the depth learning U^(2)-Net network model,which fully excavates the geometric and texture information of optical remote sensing images and conducts refined extraction of high-speed railways and their accompanying highways,using the remote sensing image satellite data of GF-2 of Chengdu Chongqing Railway.The experimental results show that the U^(2)-Net network model can accurately identify and extract high-speed railways and their accompanying highways with IoU of 73.48%and accuracy of 90.2%.This shows the feasibility and reliability of using U^(2)-Net to intelligently recognize high-speed railways and their accompanying highways.The related results can provide a certain technical reference for urban planning and development and railway disaster reduction and prevention.

关 键 词：深度学习 高速铁路 伴行公路 智能识别 

分 类 号：P237[天文地球—摄影测量与遥感]