近地面成像光谱技术在水稻氮诊断方面的应用

Application of Near Surface Imaging Spectroscopy on the Diagnosis of Rice Nitrogen

传统的测量水稻氮含量的形态诊断法和化学分析法,不仅耗费了大量的人力、物力、财力,而且测量的结果也不够精确,并且得不到及时应用,很大程度上影响了水稻的生产和产量。而基于成像光谱技术对水稻进行的氮诊断,可以做到对水稻的无损坏和精确测量,降低了成本,节省了时间,氮的使用效率明显提高,同时产量也大幅度增加,做到了经济利益和生态效益的最大化。本研究是以4种氮素水平的水稻高光谱数据为数据源,经过利用平面场模型进行反射率转换和平滑等一系列的预处理,选出各类别的样本数据,提取出了水稻叶片反射光谱的敏感波段与植被指数等特征,发现最佳识别特征,从而根据识别特征建立决策识别模型,并且不断反复验证和修正模型,总体精度能够达到87.7%。

The traditional method for measuring rice nitrogen content are morphological diagnosis and chemical analysis methods,which require a lot of manpower,material resources,and financial resources.Furthermore,the results of the measurements are not accurate and cannot be applied in time,which largely affected the production and yield of rice.Based on the imaging spectral technology,rice nitrogen diagnosis can be conducted undamagedly with accurate measurement,which reduces the cost,saves time,significantly improves nitrogen use efficiency,and greatly increases yield,thus to realize the maximization of economic benefits and ecological benefits.This study uses four nitrogen levels of rice hyperspectral data as the data source.It uses the plane field model for pre-processing such as reflectivity conversion and smoothing,selects the sample data of each category,extracts the reflectance spectrum of the sensitive bands,vegetation index,and other characteristics of rice leaves,finds the best recognition feature,and establishes decision recognition model.It constantly repeats verification and correction model,the overall accuracy can reach 87.7%。