基于无人机遥感的苹果树冠层氮含量反演研究

Research on Inversion of Nitrogen Content in Apple Tree Canopy Based on Remote Sensing of Unmanned Aerial Vehicles

快速、便捷地实时获取苹果树冠层氮含量是实现精准施肥的数据基础。本研究以“秦脆”苹果树为研究对象,分别于新梢旺长期、春梢停长期、果实膨大期利用无人机遥感平台获取30、50、70 m飞行高度下的多光谱遥感图像,并同步测定冠层氮含量。从不同试验条件下的遥感图像中各提取43种植被指数,然后通过相关性分析筛选出6种敏感植被指数,利用梯度提升决策树(GBDT)算法,建立了苹果树冠层氮含量的反演模型。结果表明:GBDT算法可以在“秦脆”苹果树不同生长期的冠层氮含量反演模型建立中取得良好的效果,且降低无人机遥感试验的飞行高度可以显著提高模型的预测精度;最优模型出现在新梢旺长期30 m高度时,其R2为0.941,RMSE为0.300。本研究结果可为“秦脆”苹果树的精准施肥提供数据支撑,并为相关研究提供参考。

Rapid and convenient acquisition of real-time nitrogen content in apple tree canopy is the basis for achieving precise fertilization.In this study,using the“Qincrisp”apple trees as research objects,the multi-spectral remote sensing images were obtained by the UAV(unmanned aerial vehicles)remote sensing platform at the flight altitude of 30,50 and 70 m during the periods of new shoot growth,spring shoot stopping growth and fruit expansion,and the canopy nitrogen content was determined synchronously.Forty-three vegetation indices were extracted from the remote sensing images under different experimental conditions,and six sensitive vegetation indices were selected by correlation analysis.The inversion model of canopy nitrogen content was established by GBDT algorithm.The results showed that the GBDT algorithm could achieve better results in the establishment of canopy nitrogen content inversion models for different growth periods of apple trees.The prediction accuracy of the models could be significantly improved by reducing the flight height of the UAV.The optimal model appeared in the new shoot growth period at 30 m of flight height,and its R~2 was 0.941 and RMSE was 0.300.The results of this study could provide data support for precise fertilization of“Qincrisp”apple trees and references for related researches.