摘要
货运源头核查是进行源头治超的关键。传统依赖于台账统计的方法费时费力,相比之下,遥感技术具有大面积覆盖等优点,利用其进行货运源头核查工作可有效提高工作效率。然而,目前尚无公开可用的货运源头遥感解译标志库,故提出一种基于YOLOX的货运源头遥感解译标志库建立方法。首先,根据货运源头核查标准,提出货运源头四类场站、四类企业解译标志库建立标准,然后基于YOLOX搭建货运源头遥感图像自动检测方法,并与专家解译结果相对比。研究结果表明,用于识别铁路货场、港口、矿石以及砂石企业的AP均在0.50以上,可实现自动化建立货运源头遥感解译标志库,为利用遥感影像进行货运源头核查提供技术支撑,显著提高源头核查工作的效率。
The source verification of freight transportation is the key to conducting source overload control.The traditional method relies on ledger statistics has the shortcomings of being time-consuming and laborintensive.Compared to the method,remote sensing technology has the advantages of wide area coverage which effectively improve work efficiency for freight source verification.Among them,interpreting symbols is an important basis for analyzing and applying the source of freight transportation using remote sensing images.However,no remote sensing interpretation sample library for freight sources is publicly available currently.Thus,this paper proposed a YOLOX based method for establishing remote sensing interpretation sample library of freight sources.Firstly,based on the verification standards of freight sources,a standard for establishing the interpretation sample library for four types of freight source stations and four types of enterprises was proposed.Then,an automatic detection method for remote sensing images of freight sources was established based on YOLOX.The experimental results were compared with the results of expert interpretation.The AP values of the proposed method for identifying railway freight yards,ports,ores,and sand and gravel enterprises were all above 0.50.The results prove that the method can realize automatically establishment of a remote sensing interpretation sample library for freight sources,which provide technical support for freight source verification using remote sensing images,and significantly improve the efficiency of freight source verification work.
作者
罗卿莉
刘宇婷
蒋鑫涛
LUO Qingli;LIU Yuting;JIANG Xintao(State Key Laboratory of Precision Measurement Technology and Instruments,Tianjin University,Tianjin 300072,China)
出处
《铁道勘察》
2024年第1期1-7,15,共8页
Railway Investigation and Surveying
基金
国家自然科学基金项目(41601446)
天津市轨道交通导航定位及时空大数据技术重点实验室开放课题(TKL2023B10)
天津市自然科学基金重点项目(21JCZDJC00670)
天津市交通运输科技发展计划项目(2020-02,2022-40)
城市轨道交通数字化建设与测评技术国家工程实验室开放课题(2021ZH04)。
关键词
遥感影像
货运源头
解译标志库
深度学习
remote sensing
freight source
interpretation sample library
deep learning