扫描电镜/能谱法对粉底液的检测分类研究
Study on Detection and Classification of Liquid Foundation by SEM/EDS
倪昕蕾 1李春宇 1孔维刚2
作者信息
- 1. 中国人民公安大学侦查学院, 北京 100038
- 2. 郑州市公安局刑事科学技术研究所, 河南 郑州 450000
- 折叠
摘要
为建立一种将扫描电镜/能谱法与多元统计学结合检测粉底液物证的方法.对收集到的 50 个粉底液样本进行前处理,利用扫描电镜/能谱仪对样本进行分析测试,将 50 个样本大致分成 2 大类;结合K-Means聚类对实验数据进行处理,将 2 大类样本细化分为 4 类;最后利用随机森林算法搭建分类模型,40 个样本作为训练集,10 个样本作为测试集,测试集的预测正确率达到 86.6%,预测效果良好,可实现样本分类自动化.此方法操作简单,可实现无损检材、分类效果良好,可直接用于公安机关侦查破案,在法庭科学领域有广阔的应用前景.
Abstract
In order to establish a method that combines scanning electron microscopy/energy dispersive spectrometry with multivariate statistics to detect the physical evidence of liquid foundation,the 50 liquid foundation samples collected were pretreated,analyzed and tested by SEM/EDS.The 50 samples were roughly divided into two categories.By combining K-Means clustering to process experimental data,the two major categories of samples were further divided into four categories;Finally,a classification model was constructed using the random forest algorithm,with 40 samples as the training set and 10 samples as the testing set.The prediction accuracy of the testing set reached 86.6%,and the prediction effect was good,which can achieve sample classification automation.This method is easy to operate,can achieve non-destructive testing of materials,and has good classification effects.It can be directly used for investigating and solving cases in public security organs,and has broad application prospects in the field of court science.
关键词
扫描电镜/能谱仪/粉底液/主成分分析法/聚类分析/随机森林Key words
scanning electron microscope/energy dispersive spectrometer/liquid foundation/principal component analysis method/cluster analysis/random forest引用本文复制引用
出版年
2024
NETL
NSTL
维普
万方数据