摘要
全球微塑料污染的日益加剧对生态环境和人类健康构成了显著威胁。然而,现有微塑料检测技术在分析效率和准确性方面存在不足,迫切需要开发新方法以提升分析能力。通过研究光谱预处理、主成分分析(Principle Component Analysis,PCA)、支持向量机(Support Vector Machine,SVM)、线性判别分析(Linear Discriminant Analysis,LDA)、K-最近邻(K-Nearest Neighbors,KNN)以及决策树(Decision Tree,DT)等算法对微塑料拉曼光谱的智能识别效果。采集6种常见微塑料标准样品的拉曼光谱数据,通过“平滑+基线校正+归一化”的联合预处理技术,有效提高了模型的预测精度。通过PCA技术实现数据降维和模式识别,获得微塑料样本间的关键差异。多种机器学习模型的应用与性能评估结果表明,LDA模型在塑料制品的识别上表现最佳,准确率和召回率为65%,特异性高达93%,查准率为73%,F1分数为54%,证实了方法在实际微塑料检测中的可行性和准确性。采用的拉曼光谱与机器学习结合的方法在微塑料鉴别中具有显著优势,不仅提高了分析效率,还提高了鉴别精度,在塑料成分鉴定中具有广阔的应用前景。
The increasingly severe global microplastic pollution poses a significant threat to the ecological environment and human health.However,the existing microplastic detection techniques suffer from deficiencies in both analysis efficiency and accuracy,new methods need to be developed to improve analytical capabilities.In this paper,we conducted the intelligent recognition of microplastics using Raman spectroscopy combined with various algorithms including spectral preprocessing,principal component analysis(PCA),support vector machine(SVM),linear discriminant analysis(LDA),K-nearest neighbor(KNN),and decision tree(DT).The algorithms were evaluated for their effectiveness in recognizing microplastic Raman spectra.The spectra of six common microplastic standard samples were collected,and the prediction accuracy of the classifiers was effectively improved by the joint preprocessing technique of“smoothing,baseline correction and normalization”.The key differences between the microplastic samples were identified through the application of PCA technique for the purpose of data dimension reduction and pattern recognition.The evaluation of multiple machine learning classifiers demonstrated that the LDA classifier exhibited optimal performance in the identification of plastic products with the accuracy and recall rate of 65%,the specificity of 93%,the precision of 73%and the F1 score of 54%,which confirmed the feasibility and accuracy of this method in practical microplastic detection.These findings indicate that the combination of Raman spectroscopy and machine learning has significant advantages in microplastic identificationwhich enhances analytical efficiency and improves discrimination accuracy,and has broad application prospects in the identification of plastic components.
作者
洪子衿
张艺严
马静
孙振丽
杜晶晶
HONG Zijin;ZHANG Yiyan;MA Jing;SUN Zhenli;DU Jingjing(College of Environmental Science and Engineering,North China Electric Power University,Beijing,102206,China;Research Center for Eco-Environmental Sciences,Chinese Academy of Sciences,Beijing,100085,China)
出处
《中国无机分析化学》
CAS
北大核心
2024年第8期1047-1057,共11页
Chinese Journal of Inorganic Analytical Chemistry
基金
国家自然科学基金资助项目(U21A20290,21707077)。
