利用Spectrocam多光谱相机获取C-88马铃薯健康叶片和患晚疫病叶片的可见光及近红外通道的多光谱图像。综合考虑多光谱图像各通道间的相关性及其信息量,采用波段指数法选取两种叶片的特征波段,并通过欧氏距离聚类方法对所提取的特征波段进行分类。实验结果表明,用波段指数法提取多光谱图像的特征波段,能快速获得马铃薯叶片的信息,475 nm、558 nm、717 nm、750 nm、850 nm作为马铃薯健康叶片的特征波段,马铃薯患晚疫病叶片的特征波段是509 nm、620 nm、717 nm、750 nm和832 nm。采用欧氏距离法对健康和患病叶片进行识别,其识别率分别可达92.6%和92.8%。因此利用多光谱成像技术可以进行马铃薯病害的快速、准确识别,为实现马铃薯病害的田间实时在线监测提供了参考。

This experiment using Spectrocam multispectral camera captured the healthy C-88 potato leaves' and late bright leaves' multispectral image within visible and near infrared bands.Multispectral image correlation between different channels and the amount of information is considered comprehensively.Band index method was used to select the characteristics of the two kinds of leaves,and the Euclidean distance clustering method was used to classify the extracted feature band.The experimental results show that the band index method used to extract the multispectral image bands can quickly obtain the information of potato leaves.We got that 475 nm,558 nm,717 nm,750 nm,850 nm band as healthy leaves of potato characteristics,and the characteristics of the potato late blight cancer leaf wavelength were 509 nm,620 nm,717 nm,750 nm and 832 nm.The recognition rate of healthy and diseased leaves was 92.6% and 92.8% by using the Euclidean distance method.Thus,using multi-spectral imaging technique can rapidly and accurately identify the defect of potato to achieve the real-time online monitoring field of the potato diseases.