多源数据融合的燃气管道高后果区识别与分级方法  

Identification and grading method of HCAs in gas pipelines based on multi-source data fusion

作　　者：唐浩然 熊俊楠[1,2] 雍志玮 陈文杰 刘傲儒 王启盛 肖慧文 王荣康 TANG Haoran;XIONG Junnan;YONG Zhiwei;CHEN Wenjie;LIU Aoru;WANG Qisheng;XIAO Huiwen;WANG Rongkang(School of Civil Engineering and Geomatics, Southwest Petroleum University;Institute of Oil and Gas Spatial Information Engineering,Southwest Petroleum University;School of Geoscience and Technology, Southwest Petroleum University)

机构地区：[1]西南石油大学土木工程与测绘学院 [2]西南石油大学油气空间信息工程研究所 [3]西南石油大学地球科学与技术学院

出　　处：《油气储运》2024年第11期1306-1312,共7页Oil & Gas Storage and Transportation

基　　金：四川省科技厅重点研发项目“基于多源遥感数据的西藏农业干旱监测关键技术研究与应用”,2021YFQ0042。

摘　　要：【目的】传统燃气管道高后果区识别方式受限于人员熟练程度与人为主观判断影响,存在效率低及人为误差较大等问题。近年来,卫星影像数据与地理信息系统(Geographic Information System,GIS)的引入在一定程度上改善了上述问题,但单一数据依然难以满足燃气管道高后果区识别与分级对数据多样性的需求,限制了识别与分级工作的定量化与自动化程度。通过多源数据获取建筑物空间及属性信息,提出燃气管道高后果区识别与分级方法。【方法】基于无人机采集的高分辨率正射影像,构建一种优化建筑物边缘提取效果的深度学习模型SEU-Net(Squeeze-and-Excitation U Network)从正射影像中提取管道周边建筑物轮廓;在引入兴趣点(Point of Interest,POI)获取建筑物种类信息的同时,通过三维点云获取高精度数字高程模型(Digital Elevation Model,DEM)与数字表面模型(Digital Surface Model,DSM),计算建筑物高度;基于GIS的空间分析功能与属性处理能力,计算建筑物面积,整合建筑物空间与属性信息。设计了燃气管道高后果区识别与分级算法。【结果】以四川省遂宁市某燃气管道为例验证该方法,与传统识别结果相比,应用该方法识别出人工未能识别的建筑物17座,减少误识别高后果区管段0.369 km。【结论】该方法具备更高的准确度和效率,在燃气管道完整性管理领域具有很好的应用潜力。(图4,表1,参19)[Objective]Constrained by operator proficiency and subjectivity,the traditional method for identifying high-consequence areas(HCAs)in gas pipelines exhibits shortcomings such as low efficiency and nonnegligible human errors.In recent years,the integration of satellite imagery data and Geographic Information System(GIS)technology has alleviated these deficiencies to some extent.Nevertheless,meeting the requirements for data diversity in identifying and grading HCAs in gas pipelines using only one data source remains a challenge.This constraint further hampers the quantification and automation capabilities of the identification and grading process.The study introduces an approach for identifying and grading HCAs in gas pipelines leveraging spatial and attribute information of buildings based on multi-source data.[Methods]Based on high-resolution orthoimages captured by unmanned aerial vehicles(UAVs),a Squeeze-and-Excitation U network(SEU-Net)was developed as a deep learning model.This model improves the extraction of building edges to outline buildings around pipelines from the orthoimages.Points of interest(POI)were introduced to collect data on building types.Furthermore,a precise Digital Elevation Model(DEM)and Digital Surface Model(DSM)were created using 3D point clouds to assess building heights.The spatial analysis feature and attribute processing capabilities of GIS were utilized to calculate building areas and integrate spatial and attribute data.Moreover,an algorithm for identifying and grading HCAs in gas pipelines was devised.[Results]The developed approach was validated in the scenario of a gas pipeline situated in Suining,Sichuan Province.In contrast to the outcomes achieved through the conventional identification technique,this method successfully detected 17 buildings that were missed in the manual process,resulting in a decrease of 0.369 km in incorrectly identified pipeline sections with HCAs.[Conclusion]The methodology showcases greater accuracy and efficiency compared to the traditional method,hi

关 键 词：燃气管道 高后果区 多源数据 识别与分级 

分 类 号：TE88[石油与天然气工程—油气储运工程]