摘要
为提高混合物质组分识别的准确性、降低建模的复杂度,构建了一种先粗分、后细分的递进式表面增强拉曼光谱检测系统。首先,用一种特征峰判别算法对全部样本进行特征提取并建立粗分模型,以分离单组分和多组分物质。然后,联合归一化与主成分分析法完成光谱特征的自动提取,并建立多输出最小二乘支持向量机细分模型。最后,采用粒子群优化算法进行参数寻优,实现多组分样本成分的精确预测。选用罗丹明6G、耐尔蓝和结晶紫三种探针分子进行实验,结果表明,特征峰判别算法提取样本特征的正确率达到99.44%,粗分模型对90例盲测样本全部识别正确;细分模型对多组分样本识别的相关系数不小于0.995,均方根误差不大于2.67343%。该拉曼检测系统能实现样本的定性、定量检测,为药品等复杂物质的检测提供了一种有效的识别途径。
For better accuracy of mixture component identification and less complexity of modeling, a progressive detection system based on surface-enhanced Raman spectroscopy with rough classification followed by subdivision is constructed in this paper. First, a characteristic peak discrimination algorithm is used to extract features from all samples and establish a rough classification model, according to which substances are classified into single-component and multi-component ones. Then, automatic extraction of spectral characteristics is accomplished through normalization and principal component analysis. A subdivision model is developed upon a multi-output least squares support vector machine. Finally, the particle swarm optimization algorithm is employed to optimize the parameters so as to achieve an accurate prediction of the composition of multi-component samples. Experiments are carried out with Rhodamine 6 G, Nile blue and crystal violet as probe molecules. The results show that the characteristic peak discrimination algorithm extracts sample features with an accuracy of 99.44%. The rough classification model correctly identifies all 90 samples in the blind test. Moreover, the correlation coefficient of the subdivision model for multi-component sample identification is not less than 0.995 and the root mean square error is not more than 2.67343%. Enabling both qualitative and quantitative detection of samples, the Raman detection system proposed in this paper provides an effective identification method for future detection of complex substances such as drugs.
作者
白鹤轩
杨峰
李丹阳
徐溢
李顺波
陈李
Bai Hexuan;Yang Feng;Li Danyang;Xu Yi;Li Shunbo;Chen Li(Key Laboratory of Optoelectronic Technology&System,Ministry of Education,College of Optoelectronic Engineering,Chongqing University,Chonging 40044,China;Key Disciplines Laboratory of Novel Micro-Nano Devices and System Techmology,Chongqing University,Chongqing 400044,China)
出处
《光学学报》
EI
CAS
CSCD
北大核心
2021年第20期156-165,共10页
Acta Optica Sinica
基金
国家重点研发计划(2018YFB2002302)
国家自然科学基金(61971074,62071072)
重庆大学中央高校基本科研业务费专项(2019CDYGYB003)。
关键词
表面光学
表面增强拉曼光谱
特征提取
粒子群优化
最小二乘支持向量机
surface optics
surface-enhanced Raman spectroscopy
feature extraction
particle swarm optimization
least squares support vector machine
