摘要
目的 利用FTIR结合化学计量学方法区分恰特草与其形近植物白茶、鱼腥草,为走私犯罪和公安禁毒工作中鉴别恰特草及其伪装品、易混淆品提供依据。方法 利用FTIR采集植物样品的红外光谱数据,结合化学计量学方法分析数据。选用SG卷积平滑、去基线、归一化三种数据处理方式对原始数据进行预处理,处理后的数据通过PCA进行分类和数据降维,最后选取KNN、RF、SVM三种模式识别方法构建数学模型进行分类和预测。结果 建模预测结果利用混淆矩阵表达,其中RF模型准确度96.51%;SVM模型准确度98.84%;KNN模型准确度93.02%。结论 本实验利用FTIR建立一种快速鉴别恰特草的数学模型,其中SVM模型正确识别率最高,各参数良好,是区别干燥恰特草及其形近植物品的最优模型。
Objective To distinguish Khat from its similar plants,such as white tea and Houttuynia cordata,so as to identity Khat and its camouflage and confusing products in smuggling crime and public security anti-drug work.Methods FTIR is used to collect the infrared spectral data of plant samples,and the data are analyzed by chemometrics.The original data were preprocessed by three data processing methods,namely,SG smoothing,baseline and normalization,then the processed data were classified and reduced by PCA,finally the mathematical model was constructed by three methods:KNN,RF and SVM for classification and prediction.Results The results were expressed by confusion matrix,and the accuracy of the RF model is 96.51 %.The accuracy of the SVM model is 98.84 %;The accuracy of the KNN model is 93.02 %.Conclusion In this experiment,FTIR was used to establish a model for rapid identification of Khat grass,the SVM model has the highest correct recognition rate and good parameters,which can be used to distinguish dry khat from its similar plants.
作者
原昱
张博
狄谱旭
陈学国
安国策
李旭鹏
张璐
Yuan Yu;Zhang Bo;Di Puxu;Chen Xueguo;An Guoce;Li Xupeng;Zhang Lu(Criminal Investigation Police University of China,Liaoning Shenyang,110854;Tianjin Mingzheng Judicial Authentication Center,Tianjin,300000)
出处
《中国法医学杂志》
CSCD
2024年第3期315-319,326,共6页
Chinese Journal of Forensic Medicine
基金
公安部刑事技术"双十计划"重点攻关任务公刑侦(2023)4806号
公安部科技计划项目(2022YY09)。
