基于模糊熵和RPCCL的高分辨率遥感影像聚类分割  

Clustering segmentation of high-resolution remote sensing images based on fuzzy entropy and RPCCL

作　　者：王超[1] 童小华[1] 谢欢[1] 刘妙龙[1] 

机构地区：[1]同济大学测量与国土信息工程系,上海200092

出　　处：《中国科技论文在线》2009年第11期783-787,共5页

基　　金：国家高技术研究发展计划(863计划)(2009AA12Z124);高等学校博士学科点专项科研基金(20070247046);上海市青年科技启明星计划(跟踪)项目(08QH14022)

摘　　要：应用模糊熵阈值分割与受限制的次胜者受罚竞争学习算法(RPCCL),实现高分辨率遥感影像的聚类分割。首先根据遥感影像各波段的模糊熵曲线进行阈值划分,选取可能的聚类中心分量,然后将各波段的分量进行组合,形成初始聚类中心,再通过RPCCL算法确定真实的聚类个数以及聚类的中心。实验表明,该算法无需事先给出聚类个数以及初始聚类中心,可以实现非监督分类。This paper integrates the fuzzy entropy threshold segmentation method and RPCCL （rival penalized controlled competitive learning） algorithm to segment high-resolution remote sensing images. The center of clustering component for each band of the image is first chosen based on the fuzzy entropy curve of the corresponding band of the image. The initial clustering centers are then formed by combining the obtained clustering center components. The real clustering centers are determined by using the RPCCL method. The results show that without providing the number of the clustering centers before clustering operation, the proposed method can perform an unsupervised segmentation of remote sensing images.

关 键 词：遥感影像 聚类 RPCCL 模糊熵 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]