作　　者： ; ; ; ;

机构地区： 中国科学院广州地球化学研究所边缘海地质重点实验室

出　　处： 《微计算机信息》 2009年第4期124-125,131,共3页

摘　　要： 云检测是遥感图像处理的一大难点,同时也是图像预处理的必要环节。本文使用基于模糊集合理论的模糊C均值聚类法,对中分辨率成像光谱仪图像的四个波段进行聚类,以检测出云和下垫面。通过我国典型区域在不同季节不同时相的数据,结合多光谱综合法和目视解译的对比检验,证明该算法适用于区域性的云检测,能检测出一些小面积的云点像元。此外,该算法在一定程度上改善了需要根据先验知识设定阈值的传统云检测算法,是一种简易可行的非监督云检测算法,为下一步云的剔除工作奠定良好的基础。 Cloud detection is a large obstacle to remote sensing image processing and a necessary step to preprocess image all the while. In order to distinguish the cloud and land covers, the article applies fuzzy C-means （FCM） algorithm to four bands of Moderate Resolution Imaging Spectroradiometer （MODIS） which are sensitive to characteristic of cloud. To prove the good performance of the cloud detection algorithm, the article uses the MODIS images of typical areas in China in different seasons and different hours, combining with the Multi-spectral synthesis method and interpretation by eyes. It is proved that the algorithm is good at detecting the territorial cloud and some pixels with small area. What is more, the algorithm improves the performance of traditional cloud detection algorithms which depend on experiential knowledge and thresholds. The FCM algorithm is a kind of viable and unsupervised method for cloud detection and good preparation for cloud elimination.

关 键 词： 中分辨率成像光谱仪 模糊 均值 聚类 云检测

领　　域： [自动化与计算机技术] [自动化与计算机技术]