摘要
针对高光谱遥感影像的特点,本文采取一种优化分类波段组合的分级选择策略。利用扩展的属性依赖性公式定义了波段间的相似度。通过模糊聚类,得到对原始波段集合的模糊等价划分。在每个模糊等价波段组中,选择一个代表性波段或进行线性融合,完成对原始波段集合的初步降维。基于遗传算法并结合粗糙集理论,给出两项能提高遗传搜索效率的增效措施,从而对降维后的波段集合进行不一致优化分类波段组合的选择。实验结果表明,本文提出的高光谱遥感影像优化分类波段组合选择方法是非常有效的。
According to the feature of hyperspectral remote sensing images, a two-step strategy is adopted for selecting the optimal classification bands combination. By means of extended attribute dependent function, the similarity between bands is defined. By fuzzy clustering, the original set of bands is divided into some fuzzy equivalent subsets. In each equivalent subset, one band is selected or a new band is generated by linear fusion. Finally, based on the genetic algorithm and rough set theory, two measures that can improve the search efficiency are provided, and the best classification band combination is determined under the inconsistent condition. The experiment indicates that the proposed optimal classification bands combination selecting approach is quite effective.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
2000年第2期181-186,共6页
Pattern Recognition and Artificial Intelligence
关键词
遥感分类
波段组合
粗糙集
遗传算法
遥感技术
Remote Sensing Classification, Bands Combination, Rough Set, Fuzzy Set, Genetic Algorithm
