摘要
提出一种结合分层聚类和判别分析对笔迹成分进行分类检验的方法。利用激光显微共聚焦拉曼光谱仪对收集的市面上常见的130支黑色签字笔笔迹样本进行检测。对测量数据进行Savitzky-Golay卷积平滑和Z-score标准化处理,利用组间连接法、组内连接法和离差平方和法三种分层聚类方法对数据进行分类,将三种聚类方法所得分类结果作为判别依据进行判别分析,检验聚类方法的正确率。结合聚类树状图与正确率,最终选择在分类数为4时原始分类结果正确率为100%、留一交叉验证分类结果正确率为98.5%的离差平方和法,提出了适用于黑色签字笔笔迹拉曼光谱数据的分层聚类方法和判别验证方法。
A method was proposed to test and classify handwriting components by combining hierarchical clustering and discriminant analysis.A collection of 130 black roller pen samples commonly found on the marketwasdetected by laser confocal micro-Raman spectrometer.The Savitzky-Golay convolution smoothing and Z-score normalization processingwereperformed on the measured data,and the datawasclassified by three hierarchical clustering methods:between-groups linkage,within-groups linkage and ward’s linkage.The classification result obtained by the methodwasused as the discriminant basis for discriminant analysis,and the correct rate of the clustering methodwastested.Combined with the clustering tree graph and the accuracy rate,the method of ward’s linkagewasselected with 100%accuracyrateof original classification results and 98.5%accuracyrateof the Leave-One-Out-Cross-Validation classification results when the number of classificationwas4.A hierarchical clustering method and a discriminant verification method for handwriting Raman spectral data of black roller penwereproposed.
作者
马枭
王晓宾
王新承
MA Xiao;WANG Xiao-bin;WANG Xin-cheng(School of Forensic Science,People's Public Security University of China,Beijing 100038,China;School of Chemical Engineering,Beijing Institute of Petrochemical Technology,Beijing 102617,China)
出处
《化学研究与应用》
CAS
CSCD
北大核心
2020年第5期873-877,共5页
Chemical Research and Application
基金
国家重点研发计划项目(2017YFC082200101)资助
中国人民公安大学2018年基本科研业务费项目(2018JKF602)资助。
关键词
拉曼光谱法
黑色签字笔笔迹
分层聚类
判别分析
Raman spectroscopy
handwriting of black roller pen
hierarchical clustering
discriminant analysis
