斑病害在全球玉米产区均有爆发, 严重影响玉米产量与品质, 是一种常见的叶类疾病。 荧光光谱技术能够快速、 无损、 准确地反映作物生理信息, 动态检测其逆境响应规律。 以玉米为研究对象, 基于荧光光谱和生理参数（SPAD和Fv/Fm）融合分析, 探究玉米生理参数对不同程度斑病害的响应规律, 构建荧光光谱反演模型。 首先, 利用相关分析与峰值分析筛选荧光光谱的敏感波段, 采用多元散射校正（MSC）、 标准正态变量变换（SNV）、 多项式平滑（S-G）、 FD光谱一阶导数、 SD光谱二阶导数等5种预处理及MSC-SG-FD, MSC-FD-SG, SNV-SG-FD, SNV-SG-SD等4种建模组合方法, 以相关系数R2和均方根误差RMSE为模型效果评价指标, 确定荧光光谱反演生理参数模型的最优方法。 结果表明: 不同斑病害程度下荧光光谱特性的整体变化趋势一致, 但强度差异显著, 在波段600.000～800.000 nm内, 光谱反射率会出现明显的峰中心, 达到极值。 在波段900.000 nm之后, 反射率趋于平稳, 特征明显减少。 对于潜伏期叶片, SPAD与Fv/Fm的建模最优方法均为SNV-SG-FD, Rc为0.985 2和0.976 8, RMSEC为1.59和2.85。 对于早期发病叶片, SPAD的建模最优方法为SNV-SG-FD, Rc为0.949 7, RMSEC为3.79, Fv/Fm的建模最优方法为SNV-SG-SD, Rc为0.943 8, RMSEC为0.011 7。 模型预测性精度较高, 能够实现对早期斑病害玉米叶片SPAD和Fv/Fm的精准预测, 为玉米斑病害潜伏期与病害早期的生理信息监测提供参考依据。 研究结果可应用于大田作业, 提升田间精细化、 智能化管理水平, 为玉米高产、 优质、 优生提供理论依据与技术支撑。

Spot disease is a common foliar disease with outbreaks in maize production areas worldwide, seriously affecting maize yield and quality. Fluorescence spectroscopy can reflect the physiological information of crops quickly and accurately without loss, and dynamically detect its response pattern to adversity. In this study, we investigated the response patterns of maize physiological parameters to different degrees of spot diseases based on the fusion analysis of fluorescence spectra and physiological parameters (SPAD and Fv/Fm) and constructed a fluorescence spectral inversion model. Firstly, the sensitive bands of fluorescence spectra were screened by correlation analysis and peak analysis, and multivariate scattering correction (MSC), standard normal variable transformation (SNV), polynomial smoothing (Smoothing), and the inversion model were used. Savitzky-Golaay (S-G), FD spectral first-order derivative, SD spectral second-order derivative, and four modeling combinations such as MSC-SG-FD, MSC-FD-SG, SNV-SG-FD, SNV-SG-SD, etc. The correlation coefficient R2 and the root mean square error RMSE were used as the evaluation indexes to determine the optimal method for fluorescence spectral inversion. The results showed that modeling the different spot disease levels was not as effective as modeling the physiological parameters. The results showed that the overall trend of fluorescence spectral properties under different spot disease degrees was consistent, but the intensity varied significantly, and the spectral reflectance would show an obvious peak center and reach the extreme value in the band 600.000～800.000 nm. After the band 900.000 nm, the reflectance leveled off and the features decreased significantly. For latent phase leaves, the modeling optimal method for both SPAD and Fv/Fm is SNV-SG-FD with Rc of 0.985 2 and 0.976 8 and RMSEP of 1.59 and 0.015 0. For early onset leaves, the modeling optimal method for SPAD is SNV-SG-FD with Rc of 0.949 7 and RMSEP of 3.79, and the Fv/Fm The modeling optimal method was SNV-SG-SD with Rc of 0.943 8 and RMSEP of 0.011 7. The high predictive accuracy of the model indicates that accurate prediction of SPAD and Fv/Fm for early spot diseased maize leaves can be achieved, providing a reference basis for monitoring physiological information during the latent and early disease stages of maize spot disease. The results of this paper can be applied to field operations, which improves the level of fine and intelligent management in the field and provides the theoretical basis and technical support for high yield, high quality and eugenics of maize.