Abstract: Traditional fog detection methods based on remote sensing mainly use polar-orbiting satellite images（such as MODIS and AVHRR） to establish fog detection model. Although the spectral information of polar-orbiting satellite is richer than geostationary satellite, the transit time is usually relatively late in the day（MODIS is about 11 o’clock and AVHRR is about 14 o’clock） and the time resolution of single polar-orbiting satellite is about one day which is unable to meet the requirements of operational fog detection. The fog detection research on stationary satellite is relatively little and the main idea is to use spectral differences to detect fog. This paper chooses FY2E image to research on day fog detection based on object-oriented approach. Streamer radiative transfer model is used to simulate the 5 bands of FY2E image for fog, cloud, ice and surface combined with measured data, six fog detection characteristic parameters are established. Mean Shift algorithm is used to segment the FY2E image and regional characteristics are established based on fog detection characteristic parameters. The fog detection model is built and experimented on FY2E data. Experimental results show that the proposed fog detection in this paper achieves good results.

Key words: fog, FY2E, Streamer, object-oriented, Mean Shift

摘要： 传统基于遥感的雾检测方法，主要利用极轨卫星遥感影像（如MODIS、AVHRR）建立雾检测模型，尽管极轨卫星影像的光谱信息丰富，但是其过境时间一般比较迟，MODIS为当地时间11点左右，AVHRR为当地时间14时左右，而且单颗极轨卫星的时间分辨率为一天左右，无法很好地满足雾检测的实际需要。针对静止卫星的雾检测研究还比较少，其主要思想是利用光谱差异进行雾检测。选择FY2E影像并结合面向对象思想开展白天雾检测研究，利用Streamer辐射传输模型针对云、雾、雪、地表对FY2E影像的5个波段进行模拟，构建雪检测指数（SDI）、雾检测指数（FDI）等6个雾检测特征参数；选择特征波段，利用Mean Shift分割方法完成FY2E影像的分割，基于雾检测特征参数构建雾检测区域特征参数；建立白天雾检测模型，选择案例数据进行实验，并利用地面站点实测数据进行精度评定。实验结果表明，提出的雾检测模型取得了很好的检测效果。

关键词: 雾, FY2E, Streamer, 面向对象, Mean Shift