为了挖掘有利于光谱分类的有效特征信息，选取园艺作物黄瓜的霜霉病害为研究对象，利用窄带滤光片型多光谱成像系统，获取患病黄瓜叶面的14个可见光通道、近红外通道和全色通道的多光谱图像。在此基础上，实验采用多光谱图像的平均灰度值和标准差，通过图像可识别度模型，对6个霜霉病害样本的16通道多光谱数据可识别度排序，目的是探讨能有效识别黄瓜植株霜霉病害的特征波段。实验结果表明，用亮度信息提取霜霉病害的多光谱图像特征波段，能快速提取病害的特征波段信息。其中700，600，589和546 nm的窄带多光谱图像具有较丰富的黑白强度分布和较好的明暗对比度，可以作为黄瓜霜霉病害光谱分类的有效特征信息通道。

In order to extract spectral feature for achieving spectral classification, 16 multispectral images of pseudoperonospora cubensis for cucumber disease, including visual light band, near-infrared band and a panchromatic band, were captured for 6 samples in this paper. Then, the distinguish ability of all multispectral images were computed by average gray, standard error and background gray. We ordered the 16 bands using the distinguish ability to 6 samples. The bands with high distinguish ability were extracted as spectral feature bands of cucumber′ pseudoperonospora cubensis. The result showed that it was speedy to extract spectral feature of cucumber disease by the gray of images. The multispectral images of 700, 600, 589 and 546 nm bands have plentiful black-white distribution and good contrast. They were spectral feature bands of cucumber′ pseudoperonospora cubensis.