便携式中波近红外光谱仪在线无损检测生鲜猪肉胆固醇

发布时间：2018-07-05 19:36

本文选题：便携式近红外光谱仪 + 中波近红外光谱　；参考：《光谱学与光谱分析》2017年06期

【摘要】：应用便携式近红外光谱仪采集320份生鲜猪肉在近红外光谱中波区的光谱信息,采用不同优化方法建立猪肉胆固醇预测模型,并对异常样品的剔除及组合预处理方法对模型性能的改善进行了分析。研究表明:通过对异常值的二次剔除,并使用SG一阶导数(savitzky-golay first derivative,SG 1st D)、SG平滑(savitzky-golay smoothing,SGS)和正交信号校正(OSC)的组合预处理方法,可获得最佳生鲜猪肉胆固醇预测模型,其参数如下:校正集相关系数(Rc)=0.9137,校正标准差(standard error of calibration,SEC)=2.5607,验证集相关系数(Rp)=0.656 7,预测标准差(standard error of prediction,SEP)=4.985 5,主因子数(principal factor,PF)=4,范围误差比(ratio of performance to standard deviation,RPD)=2.5032,相对预测标准差(relative standard error of prediction,RSEP)=8.625 4%,SEP/SEC=1.946 8,说明模型在近红外光谱中波区对猪肉胆固醇的分辨能力和预测准确度较好,通过向校正集中补充代表性样品可使模型稳健性进一步改善。对检验集样品预测值(prediction value,PV)与参比值(reference value,RV)的t检验显示二者之间无显著性差异(p0.05),检验集样品总体预测准确率为62.5%,其中50~70mg·(100g)-1区段的局部预测准确率达到91.7%,可以用于生鲜猪肉胆固醇浓度的在线快速初步定量分析。该研究将便携式近红外光谱用于在近红外中波区对生鲜猪肉及肉制品中胆固醇浓度的分析和检测,通过进一步的研究和改进,可将其应用于产品的原料分级、品质和过程控制及市售产品的抽检等。
[Abstract]:A portable near infrared spectrometer was used to collect the spectral information of 320 fresh pork in the wave region of the near infrared spectrum. The prediction model of pork cholesterol was established by different optimization methods. The elimination of abnormal samples and the improvement of the model performance were analyzed. The results showed that the two times of the abnormal value were eliminated, and the results were analyzed. Using the first derivative of SG (Savitzky-Golay first derivative, SG 1st D), SG smoothing (Savitzky-Golay smoothing, SGS) and orthogonal signal correction (OSC), the prediction model of the best fresh pork cholesterol can be obtained. The parameters are as follows: correction set phase relation and calibration standard deviation =2.5607, verify set correlation coefficient (Rp) =0.656 7, predict standard deviation (standard error of prediction, SEP) =4.985 5, principal factor number (principal factor, PF), and relative prediction standard deviation 4%, 8, It shows that the model has good resolution and prediction accuracy for pork cholesterol in the near infrared spectrum, and the robustness of the model can be further improved by adding representative samples to the corrected concentration. The t test of the sample prediction value (prediction value, PV) and the ratio (reference value, RV) of the test set shows that there is no significant difference between the two. P0.05, the overall prediction accuracy of the test set sample is 62.5%, of which the local prediction accuracy of the 50~70mg (100g) -1 section reaches 91.7%, and it can be used for the on-line fast preliminary quantitative analysis of the cholesterol concentration of fresh pork. Through further research and improvement, it can be applied to raw material grading, quality and process control, and sampling inspection of commercial products.
【作者单位】：中国肉类食品综合研究中心肉类加工技术北京市重点实验室;中国农业大学食品科学与营养工程学院教育部-北京市功能乳品实验室;
【基金】：国家“十二五”科技支撑计划项目(2014BAD04B05) 北京市农业科技项目(20150155)资助
【分类号】：O657.33;TS251.51

【相似文献】