光学学报, 2020, 40(3): 0330002, 网络出版: 2020-02-01

基于光谱相似性分析的水体致病菌种类识别方法

Recognition of Waterborne Pathogens Based on Spectral Similarity Analysis
作者单位

1中国科学院安徽光学精密机械研究所中国科学院环境光学与技术重点实验室, 安徽 合肥 230031

2中国科学技术大学, 安徽 合肥 230026

3安徽大学, 安徽 合肥 230601

4合肥学院, 安徽 合肥 230601

摘要
水体致病菌的快速识别和检测对于水质安全预警具有重要意义。以大肠埃希菌、肺炎克雷伯氏菌、金黄色葡萄球菌和鼠伤寒沙门氏菌为研究对象,对其多波长透射光谱进行测量,提出了一种基于相似学原理、余弦相似度、皮尔逊相关系数和联合相似度算法的水体细菌种类识别方法。结果表明:不同的相似度算法对不同细菌的光谱差异性的敏感度不同,相似学原理对肺炎克雷伯氏菌的识别率最高,可达98.2%;余弦相似度和皮尔逊相关系数对金黄色葡萄球菌的识别率均为100%;联合相似度算法可实现不同算法的优势互补,有效提高识别结果的可靠性与稳定性,对低浓度肺炎克雷伯氏菌、金黄色葡萄球菌、鼠伤寒沙门氏菌和大肠埃希菌的识别率分别为98.2%、100%、94.1%和91.4%,对较高浓度的上述4种细菌的识别率分别为100%、100%、100%和96%。
Abstract
Rapid recognition and detection of waterborne pathogens is of considerable significance for determining water quality and ensuring its safety. In this study, the multiwavelength transmission spectra of Escherichia coli, Klebsiella pneumoniae, Staphylococcus aureus, and Salmonella typhimurium are measured. Further, a recognition method of bacterial species in water bodies is proposed based on the principle of similarity, cosine similarity, Pearson's correlation coefficient, and joint similarity algorithm. It is found that different similarity algorithms have different sensitivities to the spectral difference of different bacteria. The principle of similarity shows the highest recognition rate for Klebsiella pneumoniae, reaching 98.2%; remarkably, the recognition rate of cosine similarity and Pearson's correlation coefficient for Staphylococcus aureus are 100%. Joint similarity algorithm can realize the complementary advantages of different algorithms and effectively improve the reliability and stability of the recognition results. The recognition rates of joint similarity algorithm for low concentrations of Klebsiella pneumoniae, Staphylococcus aureus, Salmonella typhimurium, and Escherichia coli are 98.2%, 100%, 94.1%, and 91.4%, respectively, whereas the recognition rates for higher concentrations are 100%, 100%, 100%, and 96%, respectively.
00 11