基于多元建模的皮鞋鞋底的X射线荧光光谱分类识别研究

Classification of Shoe Soles by X-ray Fluorescence Spectroscopy based on Multivariate Modeling

为建立一种快速分类及认定案件现场的皮鞋鞋底物证的分类预测模型,采用X射线荧光光谱法(X-ray Fluorescence),工作模式,电压:50 kV,电流:200μA,测试时间:60 s的实验条件下对样品进行了分析,提取了55个皮鞋鞋底样品中的无机元素成分及含量。根据提取样品所含元素种类及含量进行系统聚类,以系统聚类结果作为建立了判别分析、RBF神经网络、MLP神经网络三种分类预测模型。最终,55个样品被分为5类,经验证,多层感知器神经网络的分类准确率相对较高,为94.4%。结果表明:该方法快速无损检材,操作简单,所建立的分类预测模型可以为公安实际办案提供新思路、新方法。

In order to establish a classification and prediction model for quickly classifying and identifying the physical evidence of leather shoes and soles at the scene of the case,X-ray fluorescence spectroscopy is used and the working mode includes voltage 50 kV,current 200μA.test time 60 s.The samples were analyzed and the components and contents of inorganic elements in 55 leather shoe sole samples were extracted.According to the types and contents of elements in the extracted samples,systematic clustering was carried out.Taking the results of systematic clustering as an example,three classification and prediction models of discriminant analysis,RBF neural network and MLP neural network were established.Finally,55 samples were divided into 5 categories.After verification,the classification accuracy of multilayer perceptron neural network is relatively high,which is 94.4%.The results show that this method is fast and easy to operate,and the established classification prediction model can provide new ideas and methods for the actual handling of public security cases.