提出了基于梯度的信息散度的光谱区分方法[SID(SG)]。首先通过求取光谱梯度进行局部特征区分，再通过求光谱梯度的信息散度进行整体比较。采用仿真光谱和实际测量光谱，比较了SID(SG)与其他方法的光谱区分能力。利用相关光谱区分熵(RSDE)作为评价标准对实验结果进行了量化评价。SID(SG)方法的RSDE值分别是1.2849和1.5184，均为两组实验中几种方法的最小值。实验结果表明了SID(SG)方法相对于其他几种方法在光谱区分能力上的优越性。

A new method for spectral discrimination—spectral information divergence of spectral gradient [SID(SG)] is proposed. Firstly, the spectral gradients are estimated for discriminating the spectral local detailed characteristics, and then the information divergence of the spectral gradient is estimated for comparing their whole shape. The simulated spectra and the real measured are used as experimental data, the discrimination ability of the SID(SG) is compared to that of other methods, and relative spectral discriminatory entropy (RSDE) is used as standard to evaluate the experimental results quantitatively. RSDE values of the SID(SG) are 1.2849 and 1.5184, respectively, smaller than that of the several discrimination methods in each array. This indicates the superiority of SID(SG) over several other discrimination methods.