基于高光谱遥感的苹果树冠层叶片全氮量估测

【目的】快速准确获取大面积果园冠层叶片全氮含量（LNC ，Leaf Nitrogen Content）是实现现代精准农业的基本要求。【方法】本试验通过无人机高光谱成像仪（391.9nm ~ 1006.2nm）采集了甘肃省静宁县两个典型果园的果树冠层光谱图像，包括人工灌溉的苹果示范园与自然降雨的苹果园，综合比较两区共160份冠层叶片样本的原始光谱反射率（OD）、倒数光谱（RT）、对数光谱（LF)、一阶微分光谱（FD），构建任意两个光谱波段集组合的差值植被指数（Difference spectral index,DSI ）、土壤调节植被指数（Soil Adjusted Vegetation Index ,SAVI）、归一化光谱指数（Normalized Different Spectral Index, NDSI），分析三种光谱指数与叶片氮含量的相关性，利用一元线性回归模型与光谱指数构建两区最佳苹果冠层LNC估测模型。【结果】研究表明：人工灌溉区的FD-SAVI(825,536）、自然降雨区的LF-SAVI(854,392)与LNC的相关性最强，并基于FD-SAVI、LF-SAVI构建一元线性回归模型。人工灌溉区构建的FD-SAVI-ULRM估测模型精度最高，验证集R2和RMSE为0.6601和0.0678；自然降雨区构建的LF-SAVI-ULRM估测模型精度最高，验证集R2和RMSE为0.6746和0.0665。本试验采用LNC模型绘制出两个试验区的苹果树冠层叶片LNC估测图，实现对果园叶片全氮含量的精准掌握及精细化管理。

Estimation of total nitrogen content in canopy leaves of apple trees based on hyperspectral remote sensing

【 Objective 】 Quickly and accurately obtaining the total nitrogen content (LNC) of canopy leaves in large-scale orchards is a basic requirement for achieving modern precision agriculture. 【 Method 】 This experiment collected canopy spectral images of two typical orchards in Jingning County, Gansu Province using a drone hyperspectral imager (391.9nm~1006.2nm), including artificially irrigated apple demonstration orchards and naturally rained apple orchards. The original spectral reflectance (OD), reciprocal spectrum (RT), logarithmic spectrum (LF), and first-order differential spectrum (FD) of 160 canopy leaf samples from the two regions were comprehensively compared, Construct the Difference Spectral Index (DSI), Soil Adjusted Vegetation Index (SAVI), and Normalized Differential Spectral Index (NDSI) for any combination of two spectral bands, analyze the correlation between the three spectral indices and leaf nitrogen content, and use a univariate linear regression model and spectral indices to construct the best LNC estimation model for apple canopy in the two regions. 【 Results 】 Research shows that the correlation between FD-SAVI (825536) in artificial irrigation areas and LF-SAVI (854392) in natural rainfall areas is the strongest, and a univariate linear regression model is constructed based on FD-SAVI and LF-SAVI. The FD-SAVI-ULRM estimation model constructed in artificial irrigation areas has the highest accuracy and validation set R 2 And RMSE are 0.6601 and 0.0678; The LF-SAVI-ULRM estimation model constructed in natural rainfall areas has the highest accuracy, with validation set R 2 And RMSE is 0.6746 and 0.0665. This experiment used the LNC model to draw the LNC estimation maps of apple tree canopy leaves in two experimental areas, achieving precise control and refined management of the total nitrogen content of orchard leaves.