中文摘要：

苹果叶片氮素是反映苹果品质高低的营养元素之一。为了准确地估算苹果叶片全氮含量（LNC）,从可见光-近红外区域的高光谱反射曲线中提取光谱特征参数,应用经验回归分析,实现了对苹果LNC的高光谱监测。研究表明,除了光谱特征曲线面积变量S_（△EFG）显著相关以及面积归一化植被指数（S_（△CDE）-S_（△FGH））/（S_（△CDE）＋S_（△FGH））不相关外,其余光谱特征参数与苹果LNC都极显著相关,其中光谱特征曲线斜率K_（ge）、K_（gprv）,光谱特征曲线面积S_（△ABC）、S_（△BCD）,面积比值植被指数S_（△CDE）/S_（△ABC）、S_（△CDE）/S_（△BCD）、S_（△DEF）/S_（△ABC）,面积归一化植被指数（S_（△CDE）-S_（△ABC））/（S_（△CDE）＋S_（△ABC））、（S_（△CDE）-S_（△BCD））/（S_（△CDE）＋S_（△BCD））和（S_（△DEF）-S_（△ABC））/（S_（△DEF）＋S_（△ABC））可以较好地描述LNC的动态变化,这些特征参数对苹果LNC进行估算是可行的。通过检验,最终确定基于S_（△CDE）/S_（△ABC）、（S_（△CDE）-S_（△ABC））/（S_（△CDE）＋S_（△ABC））和（S_（△DEF）-S_（△ABC））/（S_（△DEF）＋S_（△ABC））所构建的模型为预测苹果LNC的理想模型。

英文摘要：

Apple nitrogen status is a key indicator for evaluating quality of apple fruits. In order to estimate total nitrogen content of apple leaves （LNC）, a way was proposed to monitor LNC which extracted spectral characteristics parameters from hyperspectral reflectance in the visible and near infrared regions. Hyperspectral monitoring of LNC was realized by using empirical regression analysis. Results showed that the correlation between spectral parameters and leaf nitrogen content was good in whole growth period, the best spectral parameters were Kge and S△ABC, respectively, the correlation coefficient was 0.85, the correlation between spectral parameters and leaf nitrogen content was bad, and a lot of spectral parameters were highly uncorrelated. Modeling results showed that the best model in the slope of the spectral characteristic curve was Kge of Fuji apple, the determination coefficient was 0.76, the root mean square error was 0.28, the relative error was zero, the best model in spectral characteristic curve area was S△ABC and S△BCD of gala apple, the determination coefficient was all 0.76, the root mean square error was all 0.30, the relative error was all 0.01%and zero;the best model in area ratio vegetation index was S△CDE /S△BCD and S△CDE /S△BCD of Fuji apple and S△DEF/S△ABC of Gala apple, the determination coefficient was 0.74, the root mean square error was all 0.35, the relative error was 0.01% and 0.02%, the best model in area normalized vegetation index was （S△CDE-S△BCD）/（S△CDE＋S△BCD） in the whole growth period and （S△CDE-S△ABC/（S△CDE＋S△ABC） of Gala apple, the determination coefficient was all 0.73, the root mean square error was 0.36 and 0.31, and the relative error was zero and -0.01%. The best verification results was area ratio vegetation index S△CDE/S△ABC, the determination coefficient, the root mean square error and the relative error was 0.47, 0.34 and -3.78% in the whole growth period, respectively. The determination coefficient, the ro