摘要
传统端元提取算法一般需要人工指定端元数目,易导致多选或漏选端元。提出了一种端元全自动提取方法,通过研究分类结果的均方根误差、端元间光谱相关性与端元数目的关系,设置两种循环结束条件,当分类结果的均方根误差最小或端元间恰未出现强相关现象时,获得最佳端元数。实验表明,该方法是正确有效的,能够克服现有方法的不足,提高了端元提取自动化程度。
The number of endmembers must be determinated by user conventional endmember extraction algorithm, which may result in reasonlessness of the number. Therefore, an automated endmember extraction algorithm is put forward in this paper. In it, the relationship of root mean square error of classified results, correlation between endmembers, and number of endmembers is discussed. And thus, two stop conditions, by which the most reasonable number of endmembers can be found, are given. Experimental results show that this method is proper and effective, and by it the automatization of endmember extraction will be enhanced greatly.
出处
《海洋测绘》
2009年第2期16-19,共4页
Hydrographic Surveying and Charting
关键词
端元提取
单体生长算法
循环结束条件
最佳端元数
endmember extraction
simplex growing algorithm
stop conditions
reasonable number of endmembers
