欢迎访问《麦类作物学报》编辑部网站

为了实现快速高精度获取冬小麦氮营养指数的高光谱监测技术，利用美国SVC HR-1024I型野外光谱辐射仪对2017-2019年关中地区的冬小麦进行遥感监测，获取“三边”参数、任意两波段光谱指数和植被指数，通过相关性分析和逐步回归分析方法筛选冬小麦氮营养指数的敏感光谱参数，结合偏最小二乘回归（PLSR）、随机森林算法（RFR）、支持向量机回归（SVR）和梯度增强回归（GBDT）建立冬小麦氮营养指数模型，并对模型估算精度进行验证。结果表明，从拔节期到灌浆期，各时期的氮营养指数与任意两波段光谱指数均呈极显著相关，其中拔节期氮营养指数与任意两波段光谱指数相关性均高于其他时期，且基于一阶导数光谱的归一化光谱指数和比值光谱指数与氮营养指数的相关系数最大，为0.66。拔节期基于梯度增强回归的冬小麦氮营养指数预测模型的决定系数（r²）和均方根误差（RMSE）分别为0.96和0.05，模型验证的r²、RMSE和相对预测偏差（RPD）分别为0.95、0.12和2.12，模型预测精度最高。因此，拔节期基于梯度增强回归的冬小麦氮营养指数估算模型可用于冬小麦氮营养监测及后期田间管理。

英文摘要:

It is of great significance to quickly obtain the winter wheat nitrogen nutrition index for monitoring the nitrogen nutrition status of winter wheat and guiding later fertilization. In this study, the U.S. SVC HR-1024I field spectral radiometer was used for remote sensing monitoring of winter wheat fields in Guanzhong area from 2017 to 2019 to obtain the winter wheat canopy spectrum, and the first derivative, logarithm and continuous removal spectrum were extracted to construct “trilateral parameters”, any two-band spectral indices and vegetation indices. Through correlation analysis and stepwise regression analysis of the “trilateral” parameters, any two-band spectral indices, the reported vegetation indices and the nitrogen nutrient index, the sensitive spectral parameters were screened. Combined with Partial Least Squares Regression (PLSR), Random Forest Regression (RFR), Support Vector Machine Regression (SVR), and Gradient Boosted Regression (GBDT), the nitrogen nutrition index model of winter wheat was therefore established.The results showed that the nitrogen nutrition index of each key growth stage was highly significantly correlated with the spectral indices of any two bands, among which the correlation between the nitrogen nutrition index at jointing stage and any two-band spectral indices was higher than that at other growth stages, and the correlation coefficient between the ratio spectral index based on the first derivative spectrum and the nitrogen nutrition index was the highest (0.66).The determination coefficient(r²) and root-mean-square error(RSME) of the winter wheat nitrogen nutrient index prediction model on GBDT were 0.78 and 0.87, respectively;the determination coefficient r², RMSE and relative percent deviation(RPD) of the validation model were 0.95, 0.12,and 2.12, respectively. With the highest accuracy and excellent sample prediction ability, the winter wheat nitrogen nutrient index model based on GBDT can provide technical support for the diagnosis and monitoring of winter wheat nitrogen nutrition and later field management.

查看全文查看/发表评论下载PDF阅读器

关闭