利用图像识别技术的能见度观测研究  

Research on Visibility Observation Using Image Recognition Technology

作　　者：李伟雄[1] 闫加俊 杨丽丽 LI Weixiong;YAN Jiajun;YANG Lili(Guangxi Zhuang Autonomous Region Meteorologieal and Equipment Center,Nanning 530022,China;Aerospace Newsky Technology Co.,Lud,Wuxi 214000,China)

机构地区：[1]广西壮族自治区气象技术装备中心,广西南宁530022 [2]航天新气象科技有限公司,江苏无锡214000

出　　处：《自动化仪表》2021年第6期63-67,共5页Process Automation Instrumentation

摘　　要：针对传统透射式及散射式能见度技术的局限性以及基于图像算法测量能见度的复杂性,提出了一种基于图像目标物及背景亮度值提取的能见度计算方法。通过对定点目标物的水平图像参数提取,定性、定量地分析图像的色彩、灰阶、对比度及亮度等参数,从而用目标物及背景亮度值来计算出大气消光系数,最终得出大气能见度值。通过试验,对比了图像目标物与背景亮度提取计算能见度的方法。结果表明:以视觉能见度为参考,采用图像局部亮度的方法比整体亮度计算能见度值的方法误差更小;前者误差在10%左右,后者在2%左右。与传统能见度观测方法相比,该方法原理简单、操作方便、抗干扰性强,测量得到的能见度更接近视觉能见度,对推进我国气象能见度观测现代化具有积极意义。In view of the limitations of traditional transmission and scattering visibility technologies and the complexity of measurement visibility based on image algorithm,a calculation method of visibility based on image object and background brightness value extraction was proposed.By extracting the horizontal image parameters of the fixed point object,the color,gray scale,contrast,brightness and other parameters of the image are qualitatively and quantitatively analyzed,and then the atmospheric extinction coefficient is calculated by using the brightness value of the object and the background,and finally the atmospheric visibility value is obtained.The method of extracting and calculating the visibility between the object and the background brightness is compared experimentally.The results show that the error of local luminance is less than that of global luminance,and the error of local luminance is about 10%,while that of global luminance is about 2%.Compared with the traditional visibility observation method,this method has the advantages of simple principle,convenient operation and strong anti-interference,and the measured visibility is closer to the visual visibility,which has positive significance for promoting the modernization of meteorological visibility observation in China.

关 键 词：能见度 视程 消光系数 大气透明度 亮度 对比度 视见函数 图像特征 

分 类 号：TH89[机械工程—仪器科学与技术]