基于近红外光谱的大豆水分和粗脂肪含量的快速检测

基于近红外光谱技术快速检测大豆中水分和粗脂肪含量。方法 首先采集350-2500 nm光谱范围的大豆近红外光谱，采用光谱-理化值共生距离（SPXY）算法将大豆样本划分为校正集样本与测试集样本，然后对原始光谱分别采用多元散射校正(MSC)、标准正态变量交换(SNV)、归一化（Nor）等9种方法进行预处理，最后使用偏最小二乘回归(PLSR)分析方法建立模型对样本进行定量分析。结果 原始光谱经过多元散射校正后建立的偏最小二乘回归模型对水分的预测精度最高，其校正集和测试集的相关系数分别为0.8964和0.9055 , 均方根误差分别为0.4211和0.5933；原始光谱经过归一化处理后建立的偏最小二乘回归模型对粗脂肪的预测精度最高，其校正集和测试集的相关系数分别为0.9084和0.9295 , 均方根误差分别为0.6897和0.6462。结论 近红外光谱（NIRS）结合预处理及偏最小二乘回归法，可以快速、准确的检测大豆水分和粗脂肪含量。

Rapid detection of moisture and crude fat content in soybean based on near infrared spectroscopy

Objective To detect moisture and crude fat content in soybean rapidly based on near infrared spectroscopy(NIRS). Methods Firstly, the near infrared spectrum of soybean in the spectral range of 350-2500 nm was collected, and the soybean samples were divided into calibration set samples and test set samples by using the sample set partitioning based on joint x-y distance(SPXY) algorithm, and then the original spectrum waspretreated by nine methods, such as multiplicativescattercorrection, standard normal variable exchange, and normalization, etc.Finally, the partial least square regression analysis method was used to establish a model for quantitative analysis of samples. Results The partial least square regression model established after the original spectrum was corrected by multiple scattering has the highest prediction accuracy for water content, and the correlation coefficients of the correction set and the test set are 0.8964 and 0.9055respectively. The root mean square error were 0.4211 and 0.5933 respectively. The partial least square regression model established after the original spectrum was corrected by normalization has the highest prediction accuracy for crude fat, and the correlation coefficients of the correction set and the test set were 0.9084 and 0.9295 respectively, the root mean square error were 0.6897 and 0.6462 respectively. Conclusion NIRS combined with pretreatment and partial least square regression can detect the content of water and crude fat in soybean rapidly and accurately.