Abstract: To address the problem of nighttime cloud detection, based on the geostationary meteorological satellite Himawari-8 image data, we analyze the spectral characteristics of cloud image elements and image features, and propose a geostationary satellite nighttime cloud detection method that integrates spectral thresholding and image technology, realizing the rapid and accurate detection of geostationary satellite nighttime clouds. Using MODIS cloud products and CALIPSO radar data, the cloud detection results are qualitatively analyzed and quantitatively verified. The results show that:① The cloud detection results are basically consistent with the distribution of MODIS's cloud product MYD06;② The algorithm's average cloud detection accuracy at night reaches 80.3%;③ The cloud detection accuracy at night in different seasons varies more obviously with the seasons, and reaches up to 83.3% in summer, which can distinguish between cloudy and non-cloudy regions at night in different seasons. Therefore, the geostationary satellite nighttime cloud detection method that integrates spectral thresholding and image technology can effectively realize nighttime cloud detection and provides a new idea for nighttime cloud detection applications.

Key words: nighttime cloud detection, geostationary satellite, spectral features, image techniques