基于无人机高光谱影像的冬小麦叶片氮浓度遥感估测

Estimation of nitrogen contentration in winter wheat leaves based onhyperspectral images of UAV

叶片氮浓度(LNC)是反应作物光合作用、营养状况和长势的重要指标,为精准高效地估测不同生育期冬小麦叶片氮浓度,以新冬22为研究对象,利用无人机搭载Pika L高光谱相机获取4个关键生育期冬小麦冠层反射率数据。基于波段优化算法和相关性分析筛选LNC敏感光谱指数,结合逐步回归、多元线性回归和偏最小二乘回归建立关键生育期冬小麦叶片氮浓度估测模型,并与单变量估测模型进行比较。结果表明:基于波段优化算法筛选的组合光谱指数与LNC的相关性优于传统植被指数,且达到极显著性相关;在单变量LNC估测模型中,组合光谱指数构建的模型精度优于传统植被指数,其中,扬花期差值光谱指数(DSI(R940、R968))建立的估测模型最好,R2为0.789;多变量估测模型精度均优于单变量估测模型,其中,基于偏最小二乘回归构建的LNC估算模型最好,孕穗期和扬花期拟合效果较优,模型决定系数均为0.923,均方根误差为0.082、0.084。本研究结果可以作为冬小麦LNC估测和长势监测的科学依据。

Established leaf nitrogen concentration(LNC)is the response of crop photosynthesis,an important index of nutrition and growth.To accurately and efficiently estimate different growth period of winter wheat LNC,with the new winter 22 as the research object,using the(UAVs)Pika L hyperspectral cameras for four key growth period of winter wheat canopy reflectance data.The LNC-sensitive spectral index was screened based on the band optimization algorithm and correlation analysis.Stepwise regression,multiple linear regression,and partial least squares regression were combined to establish the estimation model of winter wheat LNC in each key growth stage,which was compared with the single variable estimation model.The results showed that(1)the correlation between the combined spectral index screened using the band optimization algorithm and LNC was stronger than that obtained using the traditional vegetation index and was extremely significant;(2)the combined spectral index in the single variable LNC estimation model allowed to obtain a more accurate model compared with the traditional vegetation index,including Yang flowering DSI(R940,R968)estimate model is set up,best R2 of 0.789.The multi-variable estimation models were more accurate than the single variable estimation models and,among them,the LNC estimation model based on partial least squares regression was the best,and the fitting effect of the booting and flowering stages was better.This model had a coefficient of determination of 0.923 and rootmean-square errors of 0.082 and 0.084.The results of this study provide a theoretical basis and technical support to estimate the LNC of winter wheat and monitor its growth.