Zhuo SUN, Key Laboratory of the Virtual Geographic Environment (Nanjing Normal University), Ministry of Education, Nanjing, Jiangsu, China., China
Min CHEN, Key Laboratory of the Virtual Geographic Environment (Nanjing Normal University), Ministry of Education, Nanjing, Jiangsu, China., China
Shishi LIU, Huazhong Agricultural University, Wuhan, Hubei, China., China
Winter oilseed rape is an important oil crop in China. To date, UAV based remote sensing is becoming more popular in agriculture. Previous research on nitrogen nutrition diagnosis of crops using multi-spectral images of UAV mainly focused on food crops, especially rice and wheat. Comparing with traditional remoting sensing images, multi-spectral images of UAV can be real-time and precise. Real-time and precise management of nitrogen is an important measure for smart agriculture of winter oilseed rape, which can increase crop yields, improve nutrients absorbing, and reduce fertilizer pollution. This research aims to support an accurate quantitative diagnosis of nitrogen content in winter oilseed rape during overwintering period. It was based on a three-year nitrogen fertilizer test of winter oilseed rape in Hubei Province between 2016 and 2019, and collected UAV multispectral images during the critical growing season. By analyzing these images and their physiological and biochemical parameters, this research established a XGBoost model, and used SHAP value to describe the characteristics. The conclusions obtained are as follows: (1) The nitrogen content of rape leaves can be accurately estimated and the estimated R2 is 0.8319. (2) The regional nitrogen application amount can be accurately estimated and the estimated R2 is 0.7163. (3) A reasonable recommended nitrogen topdressing rates can be given based on the model in (2) and the known optimal nitrogen application rate. The recommended nitrogen topdressing rates will gradually decrease with the increase of the amount of nitrogen application rate.
Mots clés : Winter Oilseed Rape|UAV Multi-spectral images|XGBoost|SHAP
A104296ZS