城市和沿海大雾遥感自动检测和云、雾分离技术研究  

Automatic Detection of Heavy Fog and Cloud-Fog Separation Technology in City and Coastal Area of Eastern China Based on Meteorological Satellite Remote Sensing Image

作　　者：周红妹 汤绪[2] 葛伟强 刘冬韡 柏桦 杨引明 

机构地区：[1]上海市卫星遥感和测量应用中心,上海201199 [2]上海市气象局,上海200030

出　　处：《高原气象》2011年第3期675-682,共8页Plateau Meteorology

基　　金：上海市科委重点课题"上海城市和沿海大雾遥感监测预警系统研究"(075115011);上海市科委重大课题"洋山港海域大风大雾实时监测和预报技术研究及应用示范"(06DZ12011)共同资助

摘　　要：鉴于大雾对沿海城市国民经济的巨大影响,迫切需要对大雾产生时的分布、范围、地点等状况进行实时监测和精细分析。本文以FY-2C/D、MTSAT静止气象卫星为主,其他极轨卫星为辅,从光谱特征出发,利用大量历史、实时同步或准同步气象卫星与地面能见度观测资料,客观分析了云、雾、下垫面的晴空象元在可见光和红外各波段的光谱特征及辐射性质,研究了大雾发生季节可见光遥感图像中下垫面反射率与云、雾反射率之间的相对关系和变化规律,提出了生成准雾区动态判别门限,进而实现准雾区自动提取方法。并在此基础上,根据雾区的结构、纹理、形态等特征,采用差分计盒维数、形态分形分析、区域跟踪检测、模板特征分析、红外特征分析等多种实用有效的雾检测和云、雾分离技术模型对雾区进行自动提取。采用以上遥感方法检测出的雾区,经地面实测资料计算验证,其雾区提取精度平均相对误差为21.03%,得到了雾区信息提取和业务应用的良好效果。Heavy fog may bring enormous disaster to the national economy in the coastal cities,so there is an urgent need for real-time monitoring and precise analysis of area,distribution and location of the fog.In this paper,a practical methodology for automatic detection of fog and cloud-fog separation technology based on meteorological satellite remote sensing images is developed,such as FY-2C/D and MTSAT,assisted by polar orbit satellite data.The characteristics of the spectral channels for the detection is outlined with consideration of the available meteorological remote sensing data,as well as the live ground visibility and some other weather observation data.Using the synchronous or quasi-synchronous observation data,the spectrum and reflective characteristics of clear,could,and fog pixels in the visible light and infrared bands of the images on the same ground surface are analyzed.The relative relationship between the reflectivity of surface,cloud and fog on the visible light band of the images is then established for the automatic detection.The method of dynamic threshold is first used to getting approximate fog area and then based on this,many powerful technologies of detecting fog are adopted for separating fog and cloud,such as differential box counting,morphology fractal analysis,zonal tracking detection,template feature analysis and infrared character analysis.The average relative error of recognized fog area by above mentioned remote sensing methods is 21.03%,which have been verified by ground observation data.The result of fog detection is good,as well as in service application.

关 键 词：气象卫星遥感 城市和沿海大雾 自动检测 云、雾分离 

分 类 号：P458.121.1[天文地球—大气科学及气象学]