花生种子含油量近红外测定模型建立

摘要

对69份花生种子样品进行近红外光谱扫描，结合索氏抽提法测定的花生种子含油量的化学值，通过多种预处理方法和回归方法建立了较精准的花生种子含油量的近红外测定模型。结果显示：经过SNV+Detrend光学处理和“2，4，4，1”数学处理的预处理以及改进偏最小二乘法（MPLS）的回归处理所建模型的效果最好，其定标相关系数（RSQ）和定标标准误差（SEC）分别为0.9787和0.2187；交叉检验相关系数（1-VR）和交叉检验标准误差（SECV）分别为0.9550和0.3201。14份验证样品的预测值和化学法测定值的相关系数（R2）为0.9354，说明所建模型可以快速准确地预测花生种子的含油量。%Near-infrared spectroscopy of peanut seed was used to determine its oil content,which was compared with the oil content data of 69 samples of peanut seeds obtained by Soxhlet extraction method. Combined with the corresponding near infrared reflectance spectroscopy,the accurate model was well established with different pretreatment methods and regression analysis. The results demonstrated that the correlation coefficients of calibration(RSQ)and the root mean square errors of calibration(SEC) through SNV、Detrend and 2,4,4,1 filter spectral pretreatment methods and MPLS regression analysis were 0.978 7 and 0.218 7 respectively. The correlation coefficients of cross-validation(1-VR)and the root mean square errors of cross-validation(SECV)were 0.955 0 and 0.320 1 respectively. The correlation coefficient(R2)between NIRS value and chemical value of 14 samples was 0.935 4,which indicated that the model could be used to predict the oil content of peanut seeds fast and exactly.