紫外可见多波长透射光谱包含了细菌微生物对光的吸收和前向散射等信息, 能反映细菌细胞的组分、 大小以及形态等特征, 具有细菌种属的特异性, 可应用于细菌微生物的快速种类鉴别。 以水体中常见细菌微生物为研究对象, 实验测量了大肠埃希氏菌、 金黄色葡萄球菌、 鼠伤寒沙门氏菌以及肺炎克雷伯菌的紫外可见多波长透射光谱, 简要分析了不同种类细菌微生物的多波长透射光谱特征; 研究了透射光谱与支持向量机多向量分析方法相结合的水体细菌微生物快速识别方法, 利用基于网格搜索法的训练集内部交叉验证获取建模所需最佳惩罚因子C和核函数参数g, 根据最优参数和LibSVM一对一多分类法建立细菌快速分类鉴别模型。 利用不同株实验细菌的透射光谱作为测试集对所建模型进行识别正确率的验证, 结果表明, 所建立的快速分类鉴别模型可以对选取的大肠埃希氏菌、 金黄色葡萄球菌、 鼠伤寒沙门氏菌以及肺炎克雷伯菌进行快速种类识别, 识别正确率为100%; 分类鉴别模型对不同大肠杆菌亚种的测试集识别正确率为100%, 证明该模型对细菌属间鉴别具有较好的稳定性。 不仅可为饮用水源细菌微生物的快速识别预警提供方法, 而且可在生物医学方面作为细菌微生物鉴别的一种简便、 快速、 准确的手段。

Multi-wavelength ultraviolet visible (UV-Vis) transmission spectra of bacteria combined the forward scattering and absorption properties of microbes, contains substantial information on size, shape, and the other chemical, physiological character of bacterial cells, has the bacterial species specificity, which can be applied to rapid species identification of bacterial microbes. Four different kinds of bacteria including Escherichia coli, Staphylococcus aureus, Salmonella typhimurium and Klebsiella pneumonia which were commonly existed in water were researched in this paper. Their multi-wavelength UV-Vis transmission spectra were measured and analyzed. The rapid identification method and model of bacteria were built which were based on support vector machine (SVM) and multi-wavelength UV-Vis transmission spectra of the bacteria. Using the internal cross validation based on grid search method of the training set for obtaining the best penalty factor C and the kernel parameter g, which the model needed. Established the bacteria fast identification model according to the optimal parameters and one-against-one classification method included in LibSVM. Using different experimental bacteria strains of transmission spectra as a test set of classification accuracy verification of the model, the analysis results showed that the bacterial rapid identification model built in this paper can identification the four kinds bacterial which chosen in this paper as the accuracy was 100%, and the model also can identified different subspecies of E. coli test set as the accuracy was 100%, proved the model had a good stability in identification bacterial species. In this paper, the research results of this study not only can provide a method for rapid identification and early warning of bacterial microbial in drinking water sources, but also can be used as the microbes identified in biomedical a simple, rapid and accurate means.