基于多分类模型的记号笔墨水红外光谱分析
|
摘要
记号笔墨水的区分鉴别在相关案件的侦破和诉讼中具有重要意义。本实验采用红外光谱法(ATR-FTIR)获取记号笔的原始光谱,并对原始光谱分别进行自动基线校正、Savitzky-Golay平滑、峰面积归一化和小波阈值去噪四种预处理消除噪声等干扰因素并确定特征波长,同时结合判别分析(DA)、径向基函数神经网络(RBF)和K近邻算法(KNN)构建分类模型。结果表明,三种模型对黑色笔的分类最准确,均实现了100%的识别,对红蓝色笔区分能力次之,相比较DA和RBF,KNN模型的分类精度最高。采用ATR-FTIR结合DA-RBF-KNN法能为记号笔的类型准确检测提供新的分析手段,且模型检测精度高,方法具有普适性和一定的借鉴意义。
It's especially important for identifying the marker pens ink in the detection and litigation of related cases. This paper acquired the original spectra of the marker pens ink by ATR-FTIR,and used automatic baseline correction,Savitzky-Golay smoothing, peak area normalization and wavelet threshold denoising to eliminate interference factors,determine the characteristic wavelength range and pretreat the original spectrum. Simultaneously,the paper combined with discriminant analysis( DA),radial basis function neural network( RBF) and K nearest neighbor algorithm( KNN) to construct multi-classification model. The results showed that the three models have the most accurate classification of black pens and achieve 100% recognition,followed by the the red and blue pens.Compared with DA and RBF,KNN model has the highest classification accuracy. Using ATR-FTIR combined with DA-RBF-KNN can provide a new analysis method for more accurate detection of marker pens ink. The method has certain universality and reference significance.
It's especially important for identifying the marker pens ink in the detection and litigation of related cases. This paper acquired the original spectra of the marker pens ink by ATR-FTIR,and used automatic baseline correction,Savitzky-Golay smoothing, peak area normalization and wavelet threshold denoising to eliminate interference factors,determine the characteristic wavelength range and pretreat the original spectrum. Simultaneously,the paper combined with discriminant analysis( DA),radial basis function neural network( RBF) and K nearest neighbor algorithm( KNN) to construct multi-classification model. The results showed that the three models have the most accurate classification of black pens and achieve 100% recognition,followed by the the red and blue pens.Compared with DA and RBF,KNN model has the highest classification accuracy. Using ATR-FTIR combined with DA-RBF-KNN can provide a new analysis method for more accurate detection of marker pens ink. The method has certain universality and reference significance.
引文
[1]牛凡,黄建同,何森等.理化检验(化学分册),2017,53(01):22~27.
[2]周欣欣,郭东东,史晓凡等.理化检验(化学分册),2015,51(10):1418~1421.
[3]孙振文,程芳彬,乔婷等.分析试验室,2017,36(12):1402~1405.
[4]白炳莲,韦珏,王海涛等.化学通报,2013,76(02):167~170.
[5]陈梓云,黄云,梁焱红等.化学试剂,2017,39(06):629~633.
[6]任淯,孙雪剑,戴晓爱等.光谱学与光谱分析,2018,38(02):352~357.
[7]王恩波,许林.多酸化学导论.北京:化学工业出版社,1998.
[8]谭琨,杜培军.光谱学与光谱分析,2008,(9):2009~2013.
[9]桑应宾.重庆大学硕士学位论文,2009.
[10]唐华,杨解君,王俊等.情报杂志,2018,37(03):64~66+70+67~69.
[11]宋丽梅,罗菁.模式识别.北京:机械工业出版社,2015:82~85.
[12]何欣龙,陈利波,王继芬等.激光与光电子学进展,2018,55(5):053001.
[2]周欣欣,郭东东,史晓凡等.理化检验(化学分册),2015,51(10):1418~1421.
[3]孙振文,程芳彬,乔婷等.分析试验室,2017,36(12):1402~1405.
[4]白炳莲,韦珏,王海涛等.化学通报,2013,76(02):167~170.
[5]陈梓云,黄云,梁焱红等.化学试剂,2017,39(06):629~633.
[6]任淯,孙雪剑,戴晓爱等.光谱学与光谱分析,2018,38(02):352~357.
[7]王恩波,许林.多酸化学导论.北京:化学工业出版社,1998.
[8]谭琨,杜培军.光谱学与光谱分析,2008,(9):2009~2013.
[9]桑应宾.重庆大学硕士学位论文,2009.
[10]唐华,杨解君,王俊等.情报杂志,2018,37(03):64~66+70+67~69.
[11]宋丽梅,罗菁.模式识别.北京:机械工业出版社,2015:82~85.
[12]何欣龙,陈利波,王继芬等.激光与光电子学进展,2018,55(5):053001.