具有部分监督的遥感影像模糊聚类方法研究及应用

遥感信息主要反映的是地球表层信息。由于地球表层系统的复杂性和开放性，地表信息是多维的、无限的，遥感信息传递过程中的局限性以及遥感信息之间的复杂相关性，决定了遥感信息其结果的不确定性和多解性。模糊分类是遥感影像分类研究的重要趋势。在Ｆｕｚｚｙ－ＩＳＯＤＡＴＡ 方法的基础上，探讨如何在样本数据集中融合部分知识和随机样本，通过聚类分析获得目标类别的模糊隶属度矩阵和特征空间的特征模式的方法。提出了基于该方法的遥感影像模糊分类模型

Fuzzy Clustering Method with Partial Supervision and Its Application on Remotely Sensed Classification

Remote sensing information is the synthetical reflection of the earth surface with definite scale space.Since the earth system is characterized by complexity and opening, the information from the earth surface also should hold the characteristics of multi dimension and infiniteness.Additionally, in the process of information transfer by remote sensing imaging system, the information is somehow attenuated or amplified, and there exists complicated correlativities among the units of remote sensing data and features.All of above reasons determine that the remote sensing geo processing and geo analysis always have uncertainty and multi solutions.Thus, fuzzy classification is one of the most important trend of research of remote sensing image classification.Based on the FUZZY ISODATA method, the article discusses the improving algorithm in which the supervised samples with priori geo knowledge and random samples can be fused.So the new method is called Fuzzy clustering algorithms with partial supervision.Finally, the method is also successfully applied in a case of land cover classification in Yuanlang plain, Hong Kong.The result shows that great improvement is arrived in precision and flexibility comparing to the ISODATA method.