基于判别分析与ANN的药品铝塑包装泡罩XRF分析

目的针对案件现场常见的药品铝塑包装泡罩,为达到对其分类识别的目的,提出系列检验分析、数据处理方法。方法采用X射线荧光光谱法对45个药品铝塑包装泡罩样本所含元素进行检验并讨论分析。对检验结果进行无监督的系统聚类,利用离差平方和法计算欧氏距离进而将未知样本分为5类。结果将分类结果作为变量进行判别分析,选取累积方差百分比为97.8%的2个判别函数,其类内平方和与总平方和之比为0.015和0.394,具有较强的解释能力。绘制的样本判别分类图将5类样本类之间相互区分开来,样本总体判别正确率为95.6%。提取样本在判别函数上的判别得分构建了人工神经网络,最终分类正确率为97.8%。结论利用X射线荧光光谱法对药品铝塑包装泡罩进行检验,将元素种类及含量作为变量进行了分类,并构建了45个药品铝塑包装泡罩样本的人工神经网络分类模型,可借助该模型进一步实现对于案件现场未知类别的药品铝塑包装泡罩样本的分类识别。

XRF Analysis of Pharmaceutical Aluminum-Plastic Packaging Blister Based on Discriminant Analysis and ANN

A series of inspection, analysis and data processing methods were proposed for common pharmaceutical aluminum-plastic packaging blister at the scene of cases in order to achieve the purpose of classification and identification. X-ray fluorescence spectrometry was used to test and analyze the elements contained in 45 pharmaceutical aluminum-plastic packaging blister samples. Unsupervised systematic clustering was performed on the test results, and the Euclidean distance was calculated by using the sum of squared deviation method to classify the unknown samples into 5 categories. The classification results were observed as discriminant analysis variables. Two discriminant functions with a cumulative variance percentage of 97.8% were selected and the Wilks'' lambda is 0.015 and 0.394, which has the strongest explanatory ability. Finally, the five types of samples were distinguished from each other, and the overall discrimination accuracy rate was 95.6%. In order to achieve the purpose of pattern recognition of samples of unknown categories, extract the discriminant score of the samples on the discriminant function to construct an artificial neural network. The final classification accuracy rate was 97.8%. X-ray fluorescence spectroscopy was used to test the pharmaceutical aluminum-plastic packaging blister, and the types and contents of elements were classified as variables and an artificial neural network classification model of 45 pharmaceutical aluminum-plastic packaging blister was constructed. This model can be used to further achieve the classification and identification of pharmaceutical aluminum-plastic packaging blister of unknown categories at the scene of cases.