基于近红外光谱多种预处理的胡杨叶片含水量预测对比

通过比较5种不同光谱预处理方法（MSC、SNV、VN、一阶导数、二阶导数）提取胡杨叶片近红外光谱信息，分别采用遗传算法(GA)和连续投影算法(SPA)筛选特征波段，建立并比较偏最小二乘回归(PLS)模型对水分含量的预测效果，研究了胡杨叶片水分含量与叶片光谱信息的关系。结果表明，基于5种预处理方法使用SPA-PLS回归模型预测的相关系数R分别为0.764 4、0.869 79、0.806 01、0.779 93、0.816 8；预测均方根误差(RMSEP)分别为0.017 87、0.014 491、0.018 547、0.020 228、0.018 089；所选取的特征波段个数分别为11、20、24、18、18，较GA-PLS选取的特征波段数少，且预测效果普遍优于GA-PLS，其中基于SNV的预测结果最好。研究表明，基于近红外光谱数据，SPA算法相比于GA算法具有更好的选择特征波长能力，并且SPA-PLS算法的回归预测结果普遍优于GA-PLS，采用SNV-SPA-PLS方法可实现胡杨叶片水分含量的快速检测。

Prediction and Comparison of Leaf Water Content of Populus euphratica Based on Multiple Pretreatment of Near Infrared Spectrum

Five different spectrum pretreatment methods,i.e.,multiplication scattering correction (MSC),standard normal variate (SNV),vector normalization (VN),first-order derivative (FOD),and second-order derivative (SOD) were used to extract near infrared spectrum information of Populus euphratica leaf.Characteristic bands were screened by using genetic algorithm (GA) and continuous projection algorithm (SPA) methods.Models which were used to predict the leaf water content of Populus euphratica were established based on the partial least-squares regression (PLS).Predicted results of the models established were compared.The relationship between leaf water content and spectrum information was investigated.The correlation coefficients (R) of SPA-PLS regression models based on five pretreatment methods (MSC,SNV,VN,FOD,and SOD) were 0.764 4,0.869 79,0.806 01,0.779 93 and 0.816 8,respectively.The RMS errors was 0.017 87,0.014 491,0.018 547,0.020 228,0.018 089,respectively.The numbers of feature bands selected were 11,20,24,18,and 18,respectively,which were less than those of GA-PLS,and the prediction results were generally better than GA-PLS,among which the prediction result based on SNV was the best.The study showed that based on near-infrared spectral data,SPA algorithm had a better ability to select characteristic wavelengths than GA algorithm,and the regression prediction results of SPA-PLS algorithm were generally superior to GA-PLS.SNV-SPA-PLS method can be used to realize the rapid detection of water content in poplar leaf.