【摘要】 采用高光谱近地遥感技术,获取不同氮素水平下的玉米冠层高光谱数据,通过相关性、线性和非线性的分析方法,探讨玉米整个生育期中各营养器官的全氮含量与多种高光谱参数的相关关系,建立全氮含量的定量估测模型,并利用第2年的数据进行精度检验。结果表明:玉米冠层高光谱数据经过微分、倒数、对数处理后与玉米全氮含量的相关性均有所提高,所选波段基本在可见光波段范围内变化,其中微分处理可以显著提高数据间相关性;在10种植被指数中,玉米叶片全氮含量与植被指数RDVI(811,743)、DVI(811,754)、MSAVI(811,754)相关性都达0.92**以上,玉米叶鞘、茎秆全氮含量分别与RDVI(952,751)、RVI(948,692)相关性达到最大r=0.7867**,r=0.8444**(n=360);植被指数RVI、NDVI与玉米各营养器官叶鞘茎的全氮含量相关性普遍较好,相关系数都在0.75**以上,适合建立玉米植株全氮含量的高光谱估测模型;经过精度检验后,选择均方根误差RMSE小,拟合优度R2大,观测值与预测值的决定系数r2大的回归方程作为估测模型,得出以DVI(811,754)为自变量,估算玉米叶片全氮含量的指数模型y=2.3035e3.6854x(R2=0.859,RMSE=0.0504,r2=0.8772);以RVI(952,743)为自变量,估算玉米鞘全氮含量的二次模型y=1.5964x2-3.5464x+2.8995(R2=0.758,RMSE=0.0261,r2=0.8005);以410nm处的二阶微分为自变量,模拟玉米茎秆全氮含量的二次模型y=7454.8x2+55.335x+0.6496(R2=0.78,RMSE=0.1376,r2=0.8287)。更多还原

【Abstract】 Maize canopy hyperspectral reflectance under different nitrogen levels was obtained through the Field Hyperspectral Remote Sensing. The correlation and linear and nonlinear methods were used to analyze the relationships between total nitrogen content of maize at the whole growth stages and several hyperspectral parameters and to build the estimation models of maize total nitrogen contents. The data of the second year were used to verify the four models’ precision. The results showed that the transformation derivative reciprocal or logarithm could generally improve the correlation between the total nitrogen content and the canopy hyperspectral reflectance. The wavelength changed in visible light range. In the ten vegetation indexes, the total nitrogen content of maize leaves was correlated with RDVI(811,743), DVI(811,754), MSAVI(811,754), and these correlations were more than 0.92**, the total nitrogen contents of maize sheath and stem were significantly correlated with RDVI(952,751) and RVI(948,692), r=0.7867**, r=0.844 4**(n=360). The correlations between RVI or NDVI and vegetative organs of maize were good, all above 0.75**, which was appropriate to build the estimation model of total nitrogen content. These regression models with minimal RMSE and maximal correlation coefficient were selected as the estimation models. Estimation model of total nitrogen content in leaves could be estimated with DVI(811,754), and the estimation equation was y=2.3035e3.6854x(R2=0.859, RMSE=0.0504,r2=0.8772). Estimation models of total nitrogen content in sheath could be estimated with RVI(952,743), and the estimation equation was y=1.5964x2-3.546 4x+2.8995(R2=0.758, RMSE=0.0261, r2=0.8005). Estimation model of total nitrogen content in stem could be estimated with the second derivative spectrum in 410 nm, and the estimation was y=7454.8x2+55.335x+0.649 6(R2=0.78,RMSE=0.1376, r2=0.8287).更多还原