Abstract Crop nitrogen (N) needs to be accurately predicted to allow farmers to effectively match the N supply to the crop N demand during crop growth in order to minimize environmental impacts as excess N could seep into the water supplies around the field. The objective of this study is to use Unmanned Aerial Vehicle (UAV) multispectral MicaSense imagery validated with ground hyperspectral measurements to predict canopy nitrogen weight (g/m2) of wheat and cornfields in Ontario. A simple linear regression was established to predict the canopy nitrogen weight from various vegetation indices (VI). Ratio Vegetation Index (RVI) performed the best out of all the tested vegetation indices, with an R 2 of 0.93 for the wheat fields and 0.83 for the corn fields. RVI estimation was also consistent throughout the growing season, which is optimal in precision agriculture. Once applied the RVI-based regression model to the UAV imagery, the best RMSE was 0.95 g/m2 for the wheat McColl field using the image of May 24th and 0.66 g/m2 for the corn Jack North field using the image of June 7th. Such information for accurately predicting nitrogen is important for farmers as it will lead to a more efficient fertilizer application program.

中文翻译：

---

从无人机图像中提取小麦和玉米田的田间冠层氮

摘要 需要准确预测作物氮 (N)，使农民能够在作物生长期间有效地将 N 供应与作物 N 需求相匹配，以最大程度地减少环境影响，因为过量的 N 可能会渗入田间的供水系统。本研究的目的是使用无人机 (UAV) 多光谱 MicaSense 图像，并通过地面高光谱测量进行验证，以预测安大略省小麦和玉米地的冠层氮重量 (g/m2)。建立了一个简单的线性回归来预测各种植被指数（VI）的冠层氮重量。比率植被指数 (RVI) 在所有测试的植被指数中表现最好，小麦田的 R 2 为 0.93，玉米田的 ​​R 2 为 0.83。RVI 估计在整个生长季节也是一致的，这是精准农业的最佳选择。将基于 RVI 的回归模型应用于无人机图像后，使用 5 月 24 日图像的小麦 McColl 田的最佳 RMSE 为 0.95 g/m2，使用 6 月 7 日图像的玉米 Jack North 田的最佳 RMSE 为 0.66 g/m2。这种准确预测氮的信息对农民很重要，因为它将导致更有效的施肥计划。