基于高光谱图像技术的固态发酵中芽孢杆菌的快速识别  

Quick Identification of Bacillus in the Solid-state Fermentation Based on Hyperspectral Imaging Technology

作　　者：邹小波[1] 申婷婷[1] 石吉勇[1] 朱瑶迪[1] 胡雪桃 周煦成 

机构地区：[1]江苏大学食品与生物工程学院,江苏镇江212013

出　　处：《现代食品科技》2016年第4期235-240,共6页Modern Food Science and Technology

基　　金：高新技术发展计划国家863项目(2011AA100807);全国优秀博士基金资助项目(200968);国家自然科学基金(61301239);新世纪优秀人才项目(NCET-11-00986);江苏省杰出青年基金(BK20130010);江苏省研究生创新基金(KYLX_1070)

摘　　要：利用高光谱图像技术结合模式识别方法,研究了镇江香醋固态发酵中产酸芽孢杆菌的快速识别方法。筛选3种芽孢杆菌为标准菌,以标准菌生长12 h的菌落为研究对象,利用高光谱成像系统采集图像:提取感兴趣区域(20×20)单菌落平均光谱共120条,并SNV预处理,采用主成分分析(PCA)从每幅图像优选3幅特征图像,并从每幅特征图像提取4个基于灰度共生矩阵的纹理特征变量;对光谱和图像纹理的特征变量均进行PCA,分别提取合适的主成分构建BP-ANN和KNN识别模型。其中,光谱模型识别效果优于图像模型,且BP-ANN光谱模型识别效果最优,对校正集和预测集样本的识别率分别为98.70%和97.78%,主成分因子数为5。研究表明,菌落内部特征是识别菌种属的关键,且利用高光谱图像技术识别细菌具有可行性,且快速简便。Hyperspectral imaging technology combined with pattern recognition methods were used to rapidly identify the types of acid-producing Bacillus in the solid-state fermentation for the production of Zhenjiang balsamic vinegar.First,three species of Bacillus were screened as standard bacteria.After 12 h of growth,the standard bacterial colonies were used as study objects and images were collected using a hyperspectral imaging system.Next,a total of 120 average spectra of an area of interest（20 × 20） in a single colony were extracted and processed by standard normal variate transform.Principal component analysis（PCA） was used to select three images with a characteristic wavelength from each image,and four texture characteristic variables were extracted from each image with a characteristic wavelength based on a gray level co-occurrence matrix.Principal component analysis（PCA） was conducted on the characteristic variables of the spectra and image texture,and appropriate principle components were extracted to construct k-nearest neighbor and back propagation-artificial neural network identification models.Among them,the identification results for the spectral models were better than those of the image models,and the optimal result was obtained from the back propagation-artificial neural network spectral model,whose identification rates of calibration set and prediction set were 98.70% and 97.78%,respectively,and the number of principal component factors was five.The study shows that the internal characteristics of the bacterial colony are important for identifying the species of a colony,and hyperspectral image technology can be used for rapid and convenient bacterial identification.

关 键 词：镇江香醋 芽孢杆菌 高光谱图像技术 菌种鉴定 光谱分析 快速识别 

分 类 号：TS264.22[轻工技术与工程—发酵工程]