论文部分内容阅读
为了快速并无损地检测成年橡胶树叶片的氮素含量,使用近红外光谱检测技术获取叶片的光谱数据,采用多元散射校正(MSC)对光谱数据预处理后,使用SPA(连续投影算法)提取光谱数据的有效波长,PCA(主成分分析法)提取光谱数据主成分,然后分别将提取的光谱数据特征值输入到线性回归模型PLS(偏最小二乘回归)、非线性回归模型BPNN(BP神经网络)和LSSVM(最小二乘支持向量机)中,得到6个现有主流模型:PCA-BPNN、PCA-PLS、PCA-LSSVM、SPA-BPNN、SPA-PLS和SPA-LSSVM。用这6个模型去预测实验样本数据,经比较发现SPA-LSSVM模型对于该组实验样本的预测效果最好,其预测相关系数Rp和预测残差均方根RMSEP分别为0.9253和0.1190。因此对于成年橡胶树氮素含量的光谱快速检测,SPA-LSSVM算法模型的性能更为突出,有较好的应用潜力。
In order to rapidly and nondestructively detect the nitrogen content of adult rubber tree leaves, the spectral data of leaves were obtained by near infrared spectroscopy (NIRS), and the spectral data were preprocessed by MSC (Multisensor Scattering Correction) PCA (principal component analysis) was used to extract the principal components of the spectral data, and then the eigenvalues of the extracted spectral data were input into the linear regression model PLS (partial least-squares regression), the nonlinear regression model BPNN (BP neural network) And LSSVM (Least Squares Support Vector Machine), six existing mainstream models are obtained: PCA-BPNN, PCA-PLS, PCA-LSSVM, SPA-BPNN, SPA- PLS and SPA-LSSVM. Using these six models to predict the experimental data, the SPA-LSSVM model was found to have the best prediction effect on the experimental data. The predicted correlation coefficient Rp and root mean square RMSEP of the prediction residual were 0.9253 and 0.1190 respectively. Therefore, for the rapid detection of the nitrogen content of adult rubber tree, the performance of the SPA-LSSVM algorithm model is more prominent and has better application potential.