基于近红外光谱技术的成年橡胶树叶片氮素含量检测

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。

Detection of nitrogen in mature rubber tree leaves using near infrared spectroscopy

For a quick non-destructive detection of nitrogen content in the mature rubber tree leaf,spectral data of the leaf were obtained using the near infirared spectroscopy technique in this paper.Two different methods—SPA (Successive projections algorithm) and PCA (Principal component analysis) were used to extract the effective wavelength of the spectral data and the principal component after a multiplicative scatter correction (MSC) of the data,respectively.Then,the extracted characteristic values of the spectral data were respectively input into a linear regression model—PLS (Partial least squares) and non-linear regression model—BPNN (BP neural network) and LSSVM (Least squares support vector machine) to develop six mainstream models—PCA-BPNN,PCA-PLS,PCA-LSSVM,SPA-BPNN,SPA-PLS and SPA-LSSVM.It was found that the experimental sample prediction effect using SPA-LSSVM model was the best for the group after comparison of the results with the six models,and the prediction correlation coefficient (Rp) and root mean square error (RMSEP) were 0.9253 and 0.1190,respectively.Therefore,in terms of rapid spectral detection of nitrogen content in the mature rubber tree leaf,the performance of SPA-LSSVM algorithm model is more excellent with a broad application potential.