食品科学 ›› 2020, Vol. 41 ›› Issue (22): 315-323.doi: 10.7506/spkx1002-6630-20191005-002

• 安全检测 • 上一篇    下一篇

基于高光谱成像技术的进口冰鲜牛肉新鲜度指标检测

孙宗保,梁黎明,闫晓静,邹小波,王天真,刘小裕,李君奎   

  1. (江苏大学食品与生物工程学院,江苏 镇江 212013)
  • 出版日期:2020-11-25 发布日期:2020-11-26
  • 基金资助:
    “十三五”国家重点研发计划重点专项(2016YFD0401104); 江苏高校优势学科建设工程项目;镇江市重点研发项目(SH2019019); 食品安全大数据技术北京市重点实验室(北京工商大学)开放课题(BTBD-2020KF09)

Detection of Freshness Indexes of Imported Chilled Beef Using Hyperspectral Imaging Technology

SUN Zongbao, LIANG Liming, YAN Xiaojing, ZOU Xiaobo, WANG Tianzhen, LIU Xiaoyu, LI Junkui   

  1. (School of Food and Biological Engineering, Jiangsu University, Zhenjiang 212013, China)
  • Online:2020-11-25 Published:2020-11-26

摘要: 为了对进口冰鲜牛肉的新鲜度进行快速无损鉴别,通过高光谱成像技术结合化学计量学方法对冰鲜牛肉的新鲜度特征指标进行检测。首先测定不同新鲜度样品的挥发性盐基氮(total volatile base nitrogen,TVB-N)、颜色参数(L*、a*、b*)、脱氧肌红蛋白、氧合肌红蛋白和高铁肌红蛋白含量变化,通过显著性分析确定新鲜度特征指标。然后利用高光谱成像技术采集样品光谱和图像信息。利用光谱数据结合偏最小二乘法、区间偏最小二乘法和竞争自适应重加权-偏最小二乘法(competitive adaptive reweighted sampling-partial least squares,CARS-PLS)对特征指标含量进行预测,并对不同模型的预测结果进行比较。结果表明,TVB-N、a*和b*值这些特征指标的最佳预测模型均为CARS-PLS。模型的rc分别为0.965 8、0.949 5、0.964 2,交叉验证均方根误差分别为1.06 mg/100 g、0.71、0.76;rp分别为0.963 7、0.949 4、0.942 3,均方根误差分别为1.12 mg/100 g、0.72、0.77。利用CARS-PLS模型结合图像处理绘制的特征指标含量可视化分布图可直观表征冰鲜牛肉在贮藏过程中新鲜度的变化趋势。研究表明利用高光谱成像技术可以实现冰鲜牛肉新鲜度指标的快速检测及其分布可视化。

关键词: 进口冰鲜牛肉;新鲜度特征指标;高光谱成像;快速检测;可视化

Abstract: In order to quickly and non-destructively identify the freshness of imported chilled beef, we applied hyperspectral imaging technology combined with chemometrics to detect the characteristic freshness indexes. Firstly, chilled beef with different storage times were evaluated for their total volatile basic nitrogen (TVN-N) content, color parameters (L*, a* and b*), and deoxymyoglobin, oxymyoglobin and metmyoglobin contents. The characteristic freshness indexes were selected from them by significance and correlation analysis. Then, hyperspectral imaging technology was used to collect spectral and image information of samples. Based on the spectral data, partial least squares, interval partial least squares, competitive adaptive reweighted sampling-partial least squares (CARS-PLS) models were developed to predict the characteristic indicators, and the prediction results from these models were compared. The results showed that TVB-N content, a* and b* were selected as the characteristic freshness indexes of chilled beef with different storage times. The best prediction model for TVB-N content, a* and b* was CARS-PLS, with correlation coefficient of calibration (rc) values of 0.965 8, 0.949 5 and 0.964 2; root mean square error of cross-validation (RMSECV) values of 1.06 mg/100 g, 0.7 and 0.76; correlation coefficient of prediction (rp) values of 0.963 7, 0.949 4 and 0.942 3; and root mean square error of prediction (RMSEP) values of 1.12 mg/100 g, 0.72 and 0.77 respectively. The CARS-PLS model combined with image processing was used to draw a visual distribution map for these characteristic indicators, which visually represented the trend of freshness of chilled beef during storage. This study proves that hyperspectral imaging technology allows the rapid detection and distribution visualization of chilled beef freshness indexes.

Key words: imported chilled beef; characteristic freshness indexes; hyperspectral imaging; rapid detection; visualization

中图分类号: