中文摘要：

【目的】以高光谱技术实现小麦含水量信息的快速、无损与准确获取,为小麦灌溉的精确管理提供科学依据。【方法】利用水氮胁迫试验条件下小麦主要生长期的导数光谱构建了16种新指数,将其与NDII、WBI以及NDWI等常用指数进行比较分析,筛选小麦叶片含水量反演最佳光谱指数,并利用其建立反演模型进行小麦含水量的遥感填图。【结果】在各指数中,FD730-955对小麦冠层叶片含水量的估测结果最佳,其估测模型（对数形式）校正决定系数（C-R2）与检验决定系数（V-R2）分别达0.749与0.742,优于NDII等常用指数;FD730-955所建模型对32个未知样的预测结果与实测值相似度较高,其回归拟合模型R2达0.763,RMSE仅为0.024,取得了良好预测结果,且对叶片含水量以及LAI值较高与较低的样本均具备良好的预测能力,可有效避免样本取值范围以及冠层郁闭度等因素对含水量估测的影响;反演模型对OMIS影像的填图结果与地面实测值拟合模型R2达0.647,RMSE仅为0.027,具有较高的反演精度。【结论】导数光谱可实现小麦冠层叶片含水量信息的准确估测,其中FD730-955系反演的优选指数。

英文摘要：

[Objective] A method for fast, non-destructive and accurately monitoring leaf water content （LWC） of wheat （Triticum aestivum L.） was improved with hyperspectra technology in this paper. [Method] The canopy leaf spectral reflectance in main growing seasons of wheat was collected under the condition of water stress experiment. Using the frist-order derivative spectra, 16 new hyperspectral indices were developed to quantify the wheat＇s leaf water content （LWC）. These indices were then compared with the commonly used hyperspectral indices including NDII, WBI and NDWI to screen out the spectral index which was sensitive to LWC for modeling. Using the inversion model, the OMIS image was calculated one pixel by one pixel, and the remote sensing mapping for LWC of wheat was accomplished. [Result] The accuracy of inversion model （logarithmic function） which built by index FD730.955 was higher than that by the hyperspectral indices commonly used, as indicated by a calibration determination coefficient （C-R2） of 0.749 and a validation determination coefficient （V-R2） of 0.742. The eatimation values of 32 samples in prediction set were close to the measured values, and R2 and RMSE of regression fitted model between two dataset were 0.763 and 0.024, respectivly. Furthere more, the prediction accuracy of FD730-955 was least sensitive to the change of LWC and LAI among all of the hyperspectra indices and therefore least affected by the range of sample values and canopy density when used to estimate the LWC of wheat. The R2 and RMSE of the fitting model for the inversion and measured values were 0.635 and 0.027, respectively, and indicated the similarity between the inversion and measured value was high. [ Conclusion] It is possible to eatimate LWC of wheat by derivative hyperspectra with a high accuracy, and FD730.955 is an optimal index for modeling.