论文部分内容阅读
使用5段移动平滑法、基线校正、光谱面积归一化、多元散射校正方法对水稻叶片可见-近红外光谱进行预处理,使用连续投影算法(SPA)进行有效波长的选取。分别基于光谱指数RVI、NDVI建立多元线性回归(MLR)模型,基于SPA有效波长建立MLR模型,基于全部波长建立主成分回归(PCR)及偏最小二乘法(PLS)回归模型。利用模型预测水稻叶片氮含量,对比发现基于SPA有效波长建立的模型的预测效果明显好于基于光谱指数RVI及NDVI建立的模型,略差于基于全部波长建立的PCR及PLS模型。基于MSC预处理光谱及SPA有效波长建立的模型预测集预测结果 r=0.794 3,RMSE=0.455 8。在水稻叶片氮含量光谱监测中使用连续投影算法进行有效波长的选取是可行的。
The visible-near infrared (IR) spectra of rice leaves were preprocessed using the 5-stage moving-smoothing method, baseline correction, normalized spectral area and multivariate scatter calibration. The effective wavelength was selected using the continuous projection algorithm (SPA). Multiple linear regression (MLR) models were established based on spectral indices RVI and NDVI respectively. The MLR model was established based on the effective wavelength of SPA, and principal component regression (PCR) and partial least squares (PLS) regression models were established based on all wavelengths. Compared with the model based on spectral index RVI and NDVI, the prediction model of the model based on SPA effective wavelength was found to be slightly worse than the PCR and PLS model based on the full wavelength. Based on the pretreatment spectrum of MSC and the effective wavelength of SPA, the predicted result of model predictive set was r = 0.794 3 and RMSE = 0.455 8. It is feasible to use continuous projection algorithm to select the effective wavelength in the spectral monitoring of nitrogen content in rice leaves.