基于手持式近红外光谱仪的化学成分预测模型构建及光谱样本采集方法研究
Establishment of prediction model for chemical components based on handheld near-infrared spectrometer and collection method of sample spectra
-
摘要:
以手持式近红外光谱仪的内部烟叶化学成分预测模型为研究对象，选取296个样本作为建模集，采用偏最小二乘法（PLS）方法建立了模型，以70个样本作为实验室标准检测的外部测试集进行验证。结果表明：验证集烟碱相关系数为0.89，均方根误差为0.28，测试集烟碱相关系数为0.92，平均绝对误差为0.26，构建的预测模型精度较高。通过6点式与单点式的样本采集方法对比，6点式采样检测的烟碱平均绝对误差与单点式相比降低0.41。因此6点检测的方式较优。 Abstract:The prediction model for tobacco chemical components of a handheld near-infrared spectrometer was studied, and by selecting 296 samples as the modeling set and using partial least square (PLS) method, a model was established. Seventy samples were used as the external test set of laboratory standard detection to validate the model. The results showed that the correlation coefficient of nicotine of the validation set was 0.89, the root mean square error was 0.28, the correlation coefficient of nicotine of the test set was 0.92, and the mean absolute error was 0.26. The accuracy of the established prediction model is high. Comparing the six-point detection method and the single-point detection method, the mean absolute error of the detection values obtained through six-point detection method decreased by 0.41 compared with the values through single-point detection method. Therefore, the six-point detection method is better.
下载: