鉴于晚疫病可对马铃薯造成毁灭性灾害, 对受晚疫病胁迫的马铃薯叶片进行了高光谱图像特征研究。 旨在探索马铃薯叶片的高光谱图象特征与晚疫病害程度的关联, 以实现准确、 快速、 无损的晚疫病诊断。 采用60片马铃薯叶片, 对其中48片采用离体方式接种晚疫病菌, 所剩12片作为对照, 染病前后连续观测7天, 得到染病和健康样本。 健康和染病样本按照染病时间和染病程度不同采用374～1　018 nm波段范围的可成像高光谱仪分别采样, 基于ENVI软件处理平台提取图像中感兴趣区的光谱信息, 并采用移动平均平滑、 导数处理、 光谱变换、 基线变换等预处理方法提高信噪比, 建立了最小二乘支持向量机（LS-SVM）的识别模型。 9个模型中, 基于原始光谱(不预处理)和光谱变换预处理后的数据所建立的模型预测效果最好, 识别率均达到了94.87%。 表明基于高光谱成像技术可以实现晚疫病胁迫下马铃薯病害程度的有效区分。

Hyperspectral imaging feature on potato leaves stressed by late blight was studied in the present paper. The experiment used 60 potato leaves. Among those 60 potato leaves, 48 leaves were vitro inoculated with pathogen of potato late blight, the rest 12 leaves were used as control samples. The leaves were observed for 7 continuous days before and after inoculated and samples including healthy and infested were acquired.　Hyperspectral data of healthy and infected potato samples of different disease severity were obtained by the hyperspectral imaging system from 374 to 1　018 nm and then extract spectral data of region of interest(ROI) from those hyperspectral data by the ENVI software. In order to improve the signal-to-noise ratio, the spectral data were preprocessed using different pretreatment methods such as moving average smoothing, normalization, derivative, baseline etc. The least squares-support vector machine(LS-SVM) models were developed based on the raw and those preprocessed data. Among the nine models, the model that used the raw data and the data after the spectroscopic transformation performed best with the discrimination of 94.87%. It was demonstrated that it is realized to determine the potato late blight disease of different disease severity using hyperspectral imaging technique.