基于端元独立性的端元数目自动获取方法

目前端元数目主要是图像处理者目视判读影像获得的经验值,往往与影像中实际典型地物种类数目不符合。根据端元数目取值不同时的分类结果均方根误差(RMSE)的大小,能够得到实际端元数,但该方法因多次分类导致计算量过大。本文在研究端元间光谱相关性与端元数目之间关系的基础上,提出一种基于端元独立性的端元数目自动获取方法。与基于RMSE的端元数目自动获取方法的对比实验表明,该方法不仅能够正确有效地获取实际端元数,提高端元提取自动化程度,而且能够大大减少运算时间。

Endmember independence-based automated determination of number of endmembers

At present, the number of endmembers is empirical value achieved by contact interpretation of imagery processor, which may not be accordant with the number of true typical factors. Although the number of endmembers can be determinated by comparing RMSE of classified results, this method requires several times of classification and large computation. Based on the study of endmember independence, endmember independence-based automated determination method of number of endmembers was developed. Experiments show that the number of endmembers can be determinated accurately and computation time also can be reduced greatly by using this method.