The error propagation law shows the abundance estimation error can be reduced and hence improve the unmixed accuracy by reducing within endmember variances and increasing between endmember variances.

• 研究结果表明：(1)端元光谱间差距越大，端元光谱内差距越小，则混合像元分解精度越高；(2)如果漏选了端元光谱会造成很大的模型误差，然而多选端元光谱却仍然是端元组分的无偏估计。