叶绿素浓度是植物生长的指示剂, 而叶片的SPAD值则可以反映植物叶绿素含量, 从而监测植物的生长状况。 本文采用可见-近红外（380～1 030 nm）高光谱成像技术可以实现南瓜叶片SPAD值的可视化, 同时根据叶片霜霉病疫情与叶绿素含量呈显著正相关进而可以快速诊断霜霉病疫情。 通过测定健康叶片和感染不同霜霉病疫情的叶片光谱曲线, 采用竞争性自适应重加权算法（CARS）进行特征波段的选择, 可以得到10条特征波段, 再结合偏最小二乘回归法（PLSR）进行南瓜叶片SPAD的预测。 结果表明, 通过对48个样本的训练, 对23个样本进行预测, 可以得到南瓜叶片SPAD较好的预测效果, 其中RC=0.918, RMSECV=3.932; RCV=0.846, RMSECV=5.254; RP=0.881, RMSEP=3.714。 根据叶片光谱特征波段与SPAD之间的线性回归方程可以计算叶片各个像素点的SPAD值, 最后采用图像处理技术可以得到南瓜叶片SPAD的可视化分布图, 同时也反映了霜霉病的感染分布, 进而判断南瓜叶片的霜霉病疫情。 该研究为监测植物生长状况及判别南瓜叶片霜霉病疫情奠定了理论基础。

Visible/near-infrared (380～1 030 nm) hyperspectral imaging technique was used to realize SPAD visualization of pumpkin leaves in the present study. Downy mildew could be diagnosed rapidly according to significant positive correlation between downy mildew epidemic and chlorophyll content. Leaves uninfected and infected with different level downy mildew were used to acquire hyperspectral images and extract spectral information. Competitive adaptive reweighted sampling (CARS) was applied to select optimal wavelengths and finally 10 optimal wavelengths were obtained. Partial least squares regression (PLSR) was employed to establish SPAD prediction model. Results showed that, through the analysis of calibration of 48 samples and prediction of 23 samples, CARS-PLSR could obtain good results with RC=0.918, RMSECV=3.932; RCV=0.846, RMSECV=5.254; RP=0.881, and RMSEP=3.714. Regression model was gained based on the relationship between SPAD and spectral of pumpkin leaves. While SPAD of each pixel was calculated with PLSR regression equation, then SPAD distribution map of pumpkin was visualized using imaging processing technology. Final downy mildew infection could be diagnosed based on SPAD distribution map. This study provided a theoretical reference for effective monitoring plant growth and downy mildew epidemic.