基于近红外光谱技术的枸杞产地溯源研究

摘要

采用近红外光谱技术对宁夏、甘肃、青海、内蒙、河北的8个不同产地40种枸杞样品进行扫描,在主成分分析基础上利用简易分类法（simple modeling of class analogy,SIMCA）模式识别原理分别建立模型。结果表明：在950～1650nm全光谱波长范围内,光谱经一阶导数（5点平滑）和矢量归一化（standard normal variate,SNV）预处理后,8个产地模型的主成分数均为3时,采用SIMCA模式识别法可以建立较为稳健的枸杞产地溯源模型;在α=5%的显著水平条件下检验模型的可靠性,8个产地校正集模型的识别率除青海为80%外,其他产地均为100%,拒绝率分别为100%、100%、97%、100%、91%、94%、97%、100%,其验证集模型的识别率均为100%,拒绝率分别为100%、100%、100%、100%、75%、88%、100%、100%。表明该方法在枸杞产地识别中具有可行性。%Forty Chinese wolfberry samples from 8 different regions of China were scanned with a near infrared（NIR） spectrophotometer.The original spectra were preprocessed by the first derivative（5 point smoothing） and standard normal variate（SNV） and were subjected to principle component analysis over the wavelength range of 950 nm to 1650 nm,in which the number of principal components was selected as 3.A robust model indicating each geographic origin of Chinese wolfberry was achieved using the SIMCA（Simple Modelling of Class Analogy） pattern recognition method and its reliability at the 5% significance level was validated.The results revealed that the model indicating samples from Qinghai exhibited a recognition rate of 80% in the calibration set,the recognition rates for other 7 cultivation regions were all 100%,and the rejection rates were 100%,100%,97%,100%,91%,94%,97% and 100%,respectively.All the recognition rates of the 8 cultivation regions were 100% in the validation set,and the rejection rates were 100%,100%,100%,100%,75%,88%,100% and 100%,respectively.These results support the feasibility of applying NIR to identify the geographic origin of Chinese berry.