【摘要】 为实现干旱胁迫下冬小麦植株含水量（PWC）的定量监测，对冬小麦进行干旱胁迫，在获取冬小麦关键生育时期的冠层光谱及植株含水量基础上，采用5种常规预处理分析方法，优化原始光谱信息并分析各预处理光谱与PWC之间的关系，并利用连续投影算法（SPA）提取冬小麦PWC的光谱特征，建立并比较基于全谱（PLSR）和光谱特征波段（SPA-MLR）的冬小麦PWC光谱监测模型。结果表明，与原始光谱相比，预处理光谱可以显著提高其与冬小麦PWC的相关性，且光谱波段570、680、740、970、1 410、1 510、2 000 nm与冬小麦PWC存在紧密关系；同时采用SPA方法提取并证实光谱区域650～800、950～1 200、1 400～1 500、1 900～2 200 nm包含冬小麦PWC的重要信息。对比PLSR和SPA-MLR这2类模型表现可知，相同预处理光谱条件下的PLSR模型优于SPA-MLR模型，其中基于全谱的PLSR模型中，一阶微分预处理（1st）光谱条件下的冬小麦PWC的PLSR模型监测效果最优（Rc~2=0.957、RMSEc=0.017、RPDc=4.788），基于特征波段的MLR模型中，MSC预处理光谱条件下模型整体表现最好（Rc~2=0.814、RMSEc=0.036、RPDc=2.243），且具有实践上的应用潜力。综上可见，光谱预处理在一定程度上可以提高与PWC相关关系和影响监测模型表现，而构建模型方法可能对模型估算精度产生更为积极的效果，同时证实高光谱技术能够敏感响应干旱胁迫条件下冬小麦植株含水量，且可以实现其准确定量监测。更多还原

【Abstract】 To achieve the quantitative monitoring of winter wheat plant water content(PWC)under drought stress,in this study,a drought stress experiment was conducted on winter wheat,and the measurements of canopy spectrum and PWC were implemented at the key growth stages of winter wheat.The raw spectrum information was optimized by using five spectral pretreatments to further analyze the relationship between the preprocessed spectra and PWC.The successive projection algorithm(SPA) was adopted to extract the spectral characteristics of winter wheat PWC,established and compared the winter wheat PWC spectral monitoring models based on the full spectra of partial least square regression(PLSR) and sensitive wavelengths of SPA-multiple linear regression(MLR).The results showed that compared to the raw spectra(R),the spectral pretreatment methods could significantly improve the relationship between the preprocessed spectra and the PWC of winter wheat.The correlation analysis showed that the spectral bands 570,680,740,970,1 410,1 510,2 000 nm had a strong relationship with winter wheat PWC.Simultaneously,SPA was utilized to extract and confirm that the spectral regions with 650-800,950-1 200,1 400-1 500,1 900-2 200 nm contained the important infornation of winter wheat PWC.Comparing the two types of models of PLSR and SPA-MLR,it could be seen that the PLSR model under the same pretreatment spectral conditions outperformed the SPA-MLR model,among the PLSR models based on the full spectrum,the PLSR model of winter wheat PWC under the first order differential pretreatment(1st) condition had the best monitoring effect(Rc~2=0.957,RMSEc=0.017,RPDc=4.788),and among the MLR models,the best performance of the models was under MSC preprocessed spectral conditions(Rc~2=0.814,RMSEc=0.036,RPDc=2.243) and it had practical application potential.In conclusion,spectral pre-treatments,to some extent,could improve the correlation between the processed spectra and the PWC of winter wheat,and affect performance of the monitoring model,while the consturction of model might have a more positive effect on the accuracy of model estimation.Furthermore,it proved that the hyperspectral technology could respond sensitively to the PWC of winter wheat under drought stress and achieve accurate quantitative monitoring.更多还原