摘要 针对食源性致病菌检测过程耗时费力的问题,应用高光谱成像技术获取了12,18,24h3个生长期的5类致病菌[金黄色葡萄球菌(S.aureus)、单增李斯特菌(LM)、致泻大肠埃希氏菌(DEC)、肠炎沙门氏菌(SE)和福氏志贺菌(S.flexneri)]的高光谱图像,利用连续投影算法(SPA)结合相关分析(CA)提取不同生长期致病菌敏感波段并建立了最小二乘支持向量机(LS-SVM)判别模型。结果表明,5类致病菌对不同波长光的反射能力不同。SPA-CA筛选出的5个敏感波段(462,498,649,853,979nm)的反射率可较好地反映5类致病菌在不同生长期的波谱特性。基于该敏感波段反射率构建的LS-SVM模型能有效地判别出S.aureus、LM和DEC 3类致病菌,而SE和S.flexneri两种属容易被互相误判,SE被误判成S.flexneri的概率为11.2%,S.flexneri被误判成SE的概率为19.9%。LS-SVM模型对5类致病菌的总体识别正确率为90.9%。综上,高光谱成像技术结合化学计量学方法具有快速诊断食源性致病菌的能力。 Aiming at the problem of foodborne pathogenic bacteria detection process is time-consuming,composed a application highlights like technology obtained 12,18,24 h3 phase 5 kinds of pathogenic bacteria[Staphylococcus aureus(S.aureus),Listeria monocytogenes(LM),Diarrheagenic e.coli(DEC),Salmonella enteritidis(SE)and Shigella spp flexneri(S.flexneri)]of hyperspectral image,using successive projections algorithm(SPA)in combination with relevant analysis(CA)to extract various growing pathogenic bacteria sensitive wave bands and established the least squares support vector machines(LSSVM)discriminant model.The results showed that the reflectivity of the five types of pathogens to different wavelengths of light was different.The reflectance of the five sensitive bands(462,498,649,853,979 nm)screened by SPA-CA could well reflect the spectral characteristics of the five types of pathogens at different growth stages.The LS-SVM model based on the reflectance of this sensitive band can effectively identify S.aureus,LM and DEC,while SE and S.flexneri are easily misjudged by each other.The probability of SE being misjudged as S.flexneri is11.2%,and the probability of S.flexneri being misjudged as SE is 19.9%.The overall recognition accuracy of LS-SVM model was90.9%for the five types of pathogenic bacteria.In conclusion,hyperspectral imaging combined with stoichiometry has the ability of rapid diagnosis of foodborne pathogens.
出处 《食品与机械》 北大核心 2021年第4期63-67,86,共6页 Food and Machinery
基金 福建省农业工程高原学科建设项目(编号:712018014)。
关键词 食源性致病菌 高光谱 最小二乘支持向量机 连续投影算法 foodborne pathogenic bacteria hyperspectral imaging least squares support vector machine successive projections algorithm
