无色透明塑料包装瓶的快速无损检验  

Rapid Non-destructive Testing of Colorless Transparent PlasticPackaging Bottles

作　　者：周贯旭 姜红 周飞翔[1] 倪婷婷 黄凯 莫旖 Zhou Guanxu;Jiang Hong;Zhou Feixiang;Ni Tingting;Huang Kai;Mo Yi(Collage of Investigation,People's Public Security University of China,Beijing,100038,China;Criminal Investigation Department,Gansu Police Vocational College,Gansu Lanzhou,730046,China;Nanjing Jianzhi Instrument and Equipment Co.,Ltd.,Jiangsu Nanjing 210049,China;Guangxi Police College,Department of Criminal Science and Technology,Guangxi Nanning 530000,China)

机构地区：[1]中国人民公安大学侦查学院,北京100038 [2]甘肃警察职业学院刑事侦查系,甘肃兰州730046 [3]南京简智仪器设备有限公司,江苏南京210049 [4]广西警察学院刑事科学技术学院,广西南宁530000

出　　处：《实验与分析》2023年第2期22-27,共6页LABOR PRAXIS

基　　金：广西公安厅专项课题(项目编号:2023GAQN107)。

摘　　要：建立一种快速无损的检验无色透明塑料包装瓶的分析方法。利用差分拉曼光谱对70个无色透明塑料包装瓶样品进行检验,根据样品差分拉曼特征峰的差异,依据样品中填料的不同,将样品划分为四类,利用多层感知器神经网络对分类结果进行分析验证,多层感知器神经网络模型的正确率为95.71%。结合K-means算法对数目较多的第I类进一步聚类分析,通过戴维森堡丁指数找到最合适的K值,样品可被分为两组。同时构建支持向量机分类模型对聚类结果进行分析验证。在对于K-means聚类结果的分析验证中,支持向量机模型可以达到97.7%的准确率。结果表明该方法简单快速,样品用量少且无损样品,可为塑料包装品的物证鉴定提供科学依据。To establish a rapid and non-destructive analysis method for inspecting plastic packaging bottles.70 colorless transparent plastic packaging bottle samples were tested using differential Raman spectroscopy.Based on the differences in differential Raman characteristic peaks of the samples and the different fillers in the samples,the samples were divided into four categories.The classification results were analyzed and verified using a multi-layer perceptron neural network.The accuracy of the multi-layer perceptron neural network model was 95.71%.By combining the K-means algorithm with the Class I clustering analysis,which has a large number of logarithms,the most suitable K-value can be found through the Davidson Bauting index.The sample can be divided into two groups.Simultaneously construct a support vector machine classification model to analyze and verify the clustering results.In the analysis and validation of K-means clustering results,the support vector machine model can achieve an accuracy of 97.7%.The results indicate that this method is simple and fast,with a small sample size and non-destructive samples,which can provide a scientific basis for the identification of physical evidence of plastic packaging products.

关 键 词：差分拉曼光谱 塑料包装瓶 K-MEANS聚类 支持向量机 

分 类 号：O433.4[机械工程—光学工程] O657.33[理学—光学] TB99[理学—物理]