被动式红外成像气体目标智能检测算法及量化研究进展  

A survey of automatic gas leakage detection and quantification based on passive infrared imaging

作　　者：王琦[1,2,3] 潘夏童 邢明玮 孙云龙 赵勇 WANG Qi;PAN Xia-tong;XING Ming-wei;SUN Yun-long;ZHAO Yong(College of Information Science and Engineering,Northeastern University,Shenyang 110819,China;State Key Laboratory of Synthetical Automation for Process Industries,Northeastern University,Shenyang 110819,China;School of Control Engineering,Northeastern University at Qinhuangdao,Qinhuangdao 066004,China)

机构地区：[1]东北大学信息科学与工程学院,沈阳110819 [2]东北大学流程工业综合自动化国家重点实验室,沈阳110819 [3]东北大学秦皇岛分校控制工程学院,河北秦皇岛066004

出　　处：《控制与决策》2023年第8期2265-2282,共18页Control and Decision

基　　金：国家自然科学基金项目(62073068,62073061);中央高校基本科研业务费专项资金项目(N2204019);河北省自然科学基金项目(F2020501040);山东省自然科学基金项目(ZR2020MF108,ZR2020MD058).

摘　　要：在“双碳目标”、安全发展的时代背景下,工业气体泄漏检测因牵涉经济资源、生态环境、生产安全而成为国内外普遍关注的重要问题.气体被动式红外成像因其动态直观、可进行非接触式大范围遥测的特点而被视为检测泄漏的有效工具.在这一技术中,气体目标智能化检测及泄漏量化是两个核心的研究热点问题,且在未来一段相当长的时间内仍会是两项挑战.鉴于此,针对这两方面的研究进行综述.首先,分析基于被动式红外成像气体检测技术的原理,探究影响成像检测结果的关键因素及其作用形式;其次,将气体智能化检测算法按图像识别、视频分类、目标检测、图像分割等不同计算机视觉任务予以分类整理;再次,分别介绍气体量化任务中柱密度、路径浓度、泄漏率三者的测量方法,并重点强调不确定性分析对量化结果的重要性;最后,对气体智能化检测及量化研究中存在的问题提供一些潜在的解决方案,并展望了气体被动式红外成像技术未来的研究方向.In the era of“dual carbon goals”and safe development,industrial gas leakage detection has become an important issue of widespread concern both domestically and internationally due to its involvement in economic resources,ecological environment,and production safety.Gas passive infrared imaging is considered as an effective tool for detecting leaks due to its dynamic and intuitive characteristics,as well as the ability to perform non-contact large-scale telemetry.In this technology,intelligent detection of gas targets and leakage quantification are two core research hotspots,and they will remain two challenges for a long time to come.In view of this,a review is conducted on these two aspects of research.Firstly,the principle of passive infrared imaging gas detection technology is analyzed,and the key factors that affect imaging detection results and their forms of action are explored.Secondly,the gas intelligent detection algorithms are classified and sorted out according to different computer vision tasks such as image recognition,video classification,object detection,and image segmentation.Thirdly,the measurement methods of column density,path concentration and leakage rate in the gas quantification task are introduced respectively,and the importance of uncertainty analysis to the quantification results is emphasized.Finally,some potential solutions are provided for the problems in the research of intelligent gas detection and quantification,and the future research directions of gas passive infrared imaging technology are prospected.

关 键 词：气体泄漏检测 被动式红外成像 计算机视觉 柱密度 路径浓度 泄漏率 

分 类 号：TN247[电子电信—物理电子学]