基于粗糙集的K——均值聚类算法在遥感影像分割中的应用  

Application of Rough Sets and K-means Clustering In Remote Sensing Image Segmentation

作　　者：邵锐[1] 巫兆聪[1] 钟世明[2] 

机构地区：[1]武汉大学遥感信息工程学院,湖北武汉430079 [2]中国科学院测量与地球物理研究所,湖北武汉430077

出　　处：《现代测绘》2005年第2期3-5,共3页Modern Surveying and Mapping

基　　金：国家自然科学基金(40201039)

摘　　要：结合粗糙集理论和K———均值聚类算法,提出一种遥感影像的粗糙聚类分割方法。根据遥感影像中特征属性的相互依赖关系,应用粗糙集理论的等价关系,求出K———均值聚类所需要的初始类的个数和均值,然后采用聚类算法对图像进行分割。实验结果表明该方法比随机选取聚类的中心点和个数减少了运算量,提高了分类精度和准确性。This paper presents a remote sensing image segmentation method based on rough set theory and K-means clustering. Using equivalence relations of attributes of remote sensing image,rough set theory offers the number and the centroids of the clusters, which initialize the K-means clustering. And then the image is segmented by K-means clustering algorithm. Compared with the process beginning by random selection of objects as the centroids of the clusters, the computational time requirement is significantly small, the possibility of mistakes in segmentation is reduced and the classification precision and accuracy is improved.

关 键 词：遥感影像 粗糙集 K-均值聚类算法 分割方法 分类精度 图像处理 影像灰度空间 

分 类 号：P237[天文地球—摄影测量与遥感]