基于改进双色模型的视程障碍类天气现象识别  

Recognition of Weather Phenomena with Low Visibility on Basis of Improved Dichromatic Atmospheric Scattering Model

作　　者：郭佳[1,2] 杨玲[1,2] 吴可军[3] 马舒庆[3] 胡树贞[1,2] 

机构地区：[1]成都信息工程学院电子工程学院,成都610225 [2]中国气象局大气探测重点开放实验室,成都610225 [3]中国气象局气象探测中心,北京100081

出　　处：《气象科技》2013年第2期236-241,共6页Meteorological Science and Technology

基　　金：公益性行业(气象)专项天气现象自动化观测技术研究(GYHY200906032)资助

摘　　要：针对各种视程障碍类天气现象,在现有双色大气散射模型的基础上,通过对大气散射光方向的修正、对矢量夹角算法的改进以及加入了湿度参数信息,提出了一种改进的双色模型算法,用于识别视程类天气现象。改进算法模型考虑了大气散射光方向的不完全一致性以及景物色度像素值在RGB空间中的三维性,因而具有更高的稳定性。该算法嵌入到视程障碍类天气现象识别系统中并计算出相应的特征值,最后通过经验阈值识别天气现象;通过对算法的改进并对哥伦比亚大学一个季度和甘肃张掖观象台2个月的数据进行处理,结果显示该算法对于雾、霾、扬沙、浮尘及沙尘暴等视程障碍类天气现象具有良好的判别效果,对上述两个地区的总体识别率分别为73.2%、90.5%Concerning various kinds of low visibility weather phenomena and based on the existing dichromatic atmospheric scattering model,a modified calculating method of dichromatic model is proposed,to range weather phenomenon via revising the atmospheric scattering direction,improving the calculating method of vector angles and adding humidity parameter information.The improved algorithm model has higher stability,considering that the scatter light direction is not all consistent and pixels have three-dimensional character in the RGB space.Embedding the algorithm into the model and computing the corresponding eigenvalues,these weather phenomena can be recognized through the empirical thresholds.By improving the calculating method and processing the seasonal data of Columbia University and two-month data of Zhangye Observatory,it is shown that the improved method is excellent in identifying various kinds of low visibility weather phenomena,such as fog,haze,dust blowing,dust floating,and sand storm,whose total identification rates for the above two regions are 73.2%,90.5%,respectively.

关 键 词：双色模型 天气现象识别 特征值 大气散射光方向 

分 类 号：P412.1[天文地球—大气科学及气象学]