基于RBFNN-FDA模型对塑料拖鞋鞋底的拉曼光谱研究  

Study on Raman Spectra of Plastic Slipper Soles Based on RBFNN-FDA Model

作　　者：田陆川 杨俊[1] 姜红[1] TIAN Luchuan;YANG Jun;JIANG Hong(Investigation Institute,People’s Public Security University of China,Beijing 100038,China)

机构地区：[1]中国人民公安大学侦查学院,北京100038

出　　处：《上海塑料》2022年第4期62-67,共6页Shanghai Plastics

基　　金：国家重点研发计划项目(2018YFC1602701);中国人民公安大学2021年度基科费重点项目(2021JKF212)。

摘　　要：为建立一种快速无损检验区分塑料拖鞋鞋底的方法,利用显微共聚焦激光拉曼光谱仪采集了43个不同来源的塑料拖鞋鞋底样本的拉曼光谱图。拉曼数据经主成分分析降维后提取特征矩阵,对得到的特征矩阵进行系统聚类,建立Fisher判别函数对系统聚类的结果进行评价。最终构建径向基函数神经网络(RBFNN)实现对样本的鉴别分类,并绘制接受者操作特征曲线用以评估诊断价值。结果表明:拉曼数据提取出的特征矩阵经系统聚类被分为4组,Fisher判别分析经交叉验证后准确率为97.7%,径向基函数神经网络的准确率为100%。该方法实现了对样本快速无损的分类及预测,模型结构准确,可以为公安实际办案提供一种新思路。In order to establish a rapid and non-destructive method for distinguishing plastic slipper soles,Raman spectra of 43 plastic slipper soles from different sources were collected by micro confocal laser Raman spectrometer.After dimension reduction of Raman data by principal component analysis,feature matrix was extracted,the obtained feature matrix was systematically clustered,and Fisher discriminant function was established to evaluate the results of systematic clustering.Finally,the radial basis function neural network(RBFNN)was constructed to identify and classify the samples,and the receiver operation characteristic curve was drawn to evaluate the diagnostic value.The results show that the feature matrix extracted from Raman data can be classified into four categories by systematic clustering.The accuracy of Fisher discriminant analysis is 97.7%after cross validation,and the accuracy of radial basis function neural network is 100%.This method realizes the rapid and non-destructive classification and prediction of samples,and the model structure is accurate,which can provide a new idea for the actual handling of public security cases.

关 键 词：拉曼光谱 塑料拖鞋鞋底 系统聚类 FISHER判别分析 径向基函数神经网络 

分 类 号：O657.37[理学—分析化学]