基于无人机高清数码影像和高光谱遥感数据反演大豆典型生育期氮平衡指数

氮平衡指数(NBI)是反映作物长势的重要指标之一.通过测量NBI可以快速监测作物氮肥盈亏状况,为农业生产和管理提供精准信息.本文以无人机高清数码影像和高光谱遥感数据,以及地面实测大豆NBI数据为基础,分析大豆从开花期到成熟期,原始光谱和红外、近红外波段的导数光谱与NBI的相关性,筛选敏感波段并计算植被指数.采用经验模型法构建NBI反演模型,通过分析验证模型的决定系数(R^2)和均方根误差(RMSE),得出最佳反演模型.结果表明:大豆NBI与导数光谱反射率的相关性好于与原始光谱反射率的相关性;本文筛选的14个植被指数中,除了导数光谱光化学植被指数与大豆NBI呈不显著相关外,其余13个植被指数与大豆NBI呈极显著相关;利用13个植被指数构建NBI反演模型,并分析模型反演精度,结果显示,利用导数光谱差值植被指数构建NBI反演模型的精度最高,R^2和RMSE分别为0.771和3.077,利用该模型生成大豆典型生育期NBI分布图,能够反映大豆的长势状况.通过多载荷无人机获取的高清数码影像和高光谱遥感数据进行NBI估算,能够实时、动态、非破坏性、快速有效地监测大豆氮素营养状况,可为大豆氮肥精确管理提供简便实用的方法.

The inversion of nitrogen balance index in typical growth period of soybean based on high definition digital image and hyperspectral data on unmanned aerial vehicles

Nitrogen balance index (NBI) is one of the important indicators for crop growth. The high and low status of nitrogen can be quickly monitored by measuring NBI, which can provide accurate information of agricultural production and management. The relationship between NBI and original spectrum and derivative spectrum of infrared and near infrared wavelength from flowering to maturity stage was analyzed based on high definition digital image and hyperspectral data on unmanned aerial vehicles. Then, the sensitive bands were selected and the vegetation indexes were calculated. The inversion models of NBI were constructed by empirical model method. The optimal inversion model was obtained by analysing the determination coefficient (R²) and the root mean square error (RMSE) of validating model. The results showed that the correlation between NBI and derivative spectral reflectance was more stronger than that between it and original spectral reflectance. All the 14 vegetation indices selected in this study, except the derivative spectral photochemical reflectance index, had significant correlation with NBI. The NBI inversion models were constructed based on those 13 vegetation indices and the accuracy was analyzed. The inversion model constructed by derivative spectral difference vegetation index had the highest accuracy, with the R² and RMSE being 0.771 and 3.077 respectively. The soybean NBI distribution maps of the whole growing stages generated by this model could reflect the soybean growth state. Estimation of NBI using the high definition digital image and hyperspectral data obtained by unmanned aerial vehicle, as shown by our results, could be a real-time, dynamic, non-destructive and effective way to monitor the nitrogen status of soybean. It’s a simple and practical method for precise management of nitrogen in soybean.