基于葵花影像中国地区夜间云识别方法研究

目的实现适用于中国地区的夜间云识别。方法根据红外波段云与非云的亮温差异,采用亮温阈值法研究适用于中国地区的夜间云识别算法。中国地形高程差异明显,其地表辐射能量也存在较大差异,影响云检测结果精度,因而提出基于三个高程阶梯的云识别方法。由于云检测结果验证缺少可见光波段,使用MODIS云数据产品和雷达数据CALIPSO分别做定性和定量验证。结果云检测区域与MODIS的MYD06云产品基本一致,CALIPSO雷达数据四个月的平均验证结果为:非云区域的提取精度约77.86%,云区域的提取精度为79.67%,将有云区域错提为非云的误差率为2.76%,而将非云区域误提取为有云区域的错误率为12.31%。结论利用日本静止气象卫星Himawari-8影像数据,根据阈值法提出的基于三个高程阶梯的云检测算法,较好地实现了适用于中国地区的夜间云识别。

Nighttime Cloud Detection Method in China with Himawari-8 Image

Objective To detect nighttime cloud in China. Methods The threshold method was used to study the nighttime cloud detection in China according to the difference of brightness and temperature between cloud and non-cloud. The difference of topographic elevation in China was obvious, and the difference of surface radiation energy also existed, which affected the accuracy of cloud detection results. Therefore, the cloud detection method based on three elevation steps was proposed. Owing to the lack of visible light at night, the results of cloud detection were qualitatively and quantitatively verified by MODIS cloud products and radar laser CALIPSO data. Results The cloud detection regions were basically consistent with the MOD06 cloud products. The average verification result of CALIPSO radar data in four seasons show that, the average accuracy of extraction clear region was about 77.86%, and the extraction precision of cloud region was 79.67%. The error of extracting cloud region as non-cloud region was about 2.76%, and the error of extracting non-cloud region as cloud area was 12.31%. Conclusion Considered with three elevation steps, the threshold methods have been implemented on cloud detection at nighttime in China by Himawari-8 image data form Japanese geostationary meteorological satellite.