基于MODIS数据的云自动化识别技术

Preparatory Research on Automatic Cloud Type Classification Technique Based on MODIS Data

在对MODIS数据进行必要的预处理和数据分析的基础上采用纹理特征分类方法,进行了云与背景以及不同云类型的分类试验。结果表明,基于灰度共生矩阵提取的纹理特征能够得到较好的分类效果,同时统计分析表明,8个纹理特征中Entropy特征能更好地描述背景类(海洋、陆地)、不同云类型、以及(从反射率上)极易混淆的积雪等类型的纹理组成;Entropy特征在可见光波段1或波段4上表现出极为相似的特征。因此,实际应用时建议选择其中之一,既可以满足精度需求又可以提高处理效率。

It is the basic work to discriminate the cloud types by use of remote sensing data in the several meteorological research fields. In this paper, the classification test between cloud and background as well as different cloud types was finished after the essential pre-processing and data analysis work were performed, at the same time the operational potential of this method was discussed. The results show that the better classification effect can be got by use of traditional classification method with texture features based on the gray co-occurrence matrix theory, based on the statistical analysis the study show that the entropy feature among eight common used textures can better describe the difference between sea, land and different cloud types especially for the snow which is frequently confused with the cloud type in the visible spectrum, at the same time, the research show that the entropy yalue show the very similar statistic characteristics in bandl and band4, so only one band between which can be selected in order to improve the processing speed.