基于PLSR方法的马铃薯叶片氮素含量机载高光谱遥感反演

PLSR-based Airborne Hyperspectral Remote Sensing Retrieval of Leaf Nitrogen Content in Potato Fields

作物氮素状况是评价土壤肥力和作物长势的重要指标,叶片氮素状况的实时无损估测对合理氮素管理、提高产量和改善品质具有重要意义。本文选择不同氮处理条件下的马铃薯作为研究对象,利用AISAEagle机载高光谱成像系统获取试验区的高光谱图像,在对图像进行精确的几何、辐射校正和反射光谱重建基础上,提取每个处理马铃薯冠层的高光谱数据。选取波长430-910nm范围内原始光谱R及其D1（R）、D2（R）、Log（1/R）、DLog（1/R）、D2Log（1/R）5种变式数据,根据田间同步采样叶片的氮素含量数据,利用偏最小二乘回归法（PLSR）构建了马铃薯叶片氮素含量的光谱预测模型,并进行全氮含量填图。结果表明：基于一阶导数光谱D1（R）的偏最小二乘回归模型的效果最优,决定系数（R2）和校正均方差（RMSEC）分别为0.82、0.38%。将该最优估算模型应用到整个高光谱图像上,得到试验区马铃薯叶片全氮分布图,图像上氮的值域为3.35%-5.95%,与地面实测结果3.59%-5.89%基本一致,且叶片全氮值的大小分布与马铃薯长势分布一致。研究结果可为研制和开发基于高光谱成像技术的马铃薯叶片氮素预测方法提供理论和技术支持。

Crop nitrogen status is the important evaluation index of crop growth status and soil nitrogen level, timely and non-destructive monitoring of leaf nitrogen status is significant to optimize nitrogen management and improve grain yield and quality. In the present study, five nitrogen （ N ） fertilizer treatments were applied to build up levels of nitrogen variation in potato fields. Relationships between canopy spectral reflectance from the AISA-Eagle airborne imaging spectrometer data after geometrical and radiometric correction and reconstruction of reflectance spectrum and nitrogen levels in potato leaves were studied. The leaves were sampled and analyzed for leaf N. Leaf N content prediction models were developed using PLSR by analyzing the raw reflectance data（R） at 430 -910nm, their first derivative D1 （R）, their second derivative D2 （ R）, their absorbance equivalent [ log （1/R） ], its first derivative [ Dlog （l/R） ] and its second derivative [ D21og （l/R） ]. The results indicated that the PLSR model with the first derivative gave the best performance with R2 = 0. 82, RMSEC = 0.38%. The best estimation model from PLSR was applied to all the potato pixels of the airborne images. The values of the leaf N distribution map ranged from 3.35%o to 5.95%, which were quite consistent with those of laboratory measurements from 3.59% to 5.89%. The distribution agreed highly with the growth status distribution. The results provide an important theoretical and technical foundation for the future research and development of hyperspectral images-based for distinguishing spatial variability in N status in potato fields.