针对目前遥感影像分类方法中存在分类知识难以获取的不足, 尝试引入形式概念分析的数据挖掘理论, 并基于族集最小覆盖理论实现概念内涵的缩减, 从而保证分类规则的简洁与无冗余性。研究选取湖北省房县作为试验区, 实现了该理论在研究区中土地利用类型分类规则的挖掘应用。基于挖掘出的分类规则构建了启发式分类器, 实验结果表明形式概念分析理论挖掘出的分类规则可信度较高, 基于挖掘出的分类规则构建的分类器相对于监督分类方法、决策树C4.5算法在分类精度上有一定优势, 从而证明了它对遥感影像分类提供一种的新方法。

In order to solve the problems that how to mine and express classification knowledge and rules in current remote sensing image classification, this paper introduces a new data mining theory of formal concept analysis, and realizes the connotation reduction of concept based on the minimum coverage of sets for ensuring the simplicity of classification rules. Meanwhile, the Fang city of Hubei province is selected to carry out the formal concept analysis theory to mine the land-use types classification rules, and construct a heuristic classifier based on the mined classification rules. The result shows that the mined classification rules have higher credibility, and the constructed classifier has higher accuracy compared with supervision classification and C4.5 algorithm, which proves that the theory of formal concept analysis provides a new method to achieve remote sensing image classification.