东北寒地水稻冠层氮素含量高光谱预测模型

为实现动态、无损监测寒地水稻氮素状况,利用高光谱成像技术,分析不同生育期水稻冠层光谱特征,借助波段自相关分析(Bands inter-correlation analysis,BICA)与主成分分析(Principal components analysis,PCA),选择特征波段构建多种植被指数.根据植被指数与氮素含量相关性,建立单变量预测模型.利用最大R2增量法(MAXR)分析全部植被指数与氮素含量定量关系,建立多变量预测模型.结果表明,从分蘖期到抽穗期,寒地水稻冠层光谱反射率在可见光波段内降低,在近红外波段内增加.基于BICA-PCA-MAXR预测模型预测精度和稳定性较基于BICA-PCA结合一元回归预测模型显著提升.研究结果可为水稻氮素含量快速检测提供地域参考,水稻精准施肥管理提供技术支持.

Hyperspectral estimation model for predicting canopy nitrogen content of rice in cold region of Northeast China

In order to realize the dynamic and non-destructive monitoring of rice nitrogen status in cold region,the hyperspectral imaging technique was used to analyze the characteristics of rice canopy spectrum at different growth stages.On the basis of the verification of the previous research results,bands inter-correlation analysis (BICA) and principal components analysis (PCA) were used to select characteristic bands to construct vegetation indices.The univariate prediction models were established according to the correlation between vegetation indices and nitrogen content.Quantitative relationship between total vegetation indices and nitrogen content was analyzed by MAXR,and the multivariate prediction models were established.The obtained models were validated,analyzed and evaluated.The results showed that,from the tillering stage to heading date,the canopy spectral reflectance of rice in cold region decreased in visible region and increased in near infrared region.The prediction accuracy and stability of the prediction models based on BICA-PCA-MAXR were significantly improved compared with the prediction models based on BICA-PCA combined with simple regression analysis.The results provided regional reference for the rapid detection of rice nitrogen content and technical support for guiding the precise fertilization management.