Study On Pattern Recognition For Blue Ballpoint Pen Ink Altered Document Based On Spectroscopy And Chromatography

Altered document,is refers to a forged document which some words or strokes were added on the original document by the perpetrators in order to change the mean of the original document.In the three major types of lawsuits in our country,especially the economic crime cases,often involves the altered documents.In our study,spectroscopy and chromatography,such as Raman spectroscopy,Ultraviolet-Visible spectroscopy?UV-Vis?,Thin-layer Chromatography?TLC?and Gas Chromatography?GC?,four methods totally,were used to examine 270 different brands and types of blue ballpoint pen collected from market,and chose the spectral and chromatographic data of 25 groups different brands,such as Chenguang,Aihao,Zhencai,zebra,et al,as the training set,used ChemPattern^TM chemometrics solution software to set up classification model of different brands based on spectroscopy and chromatography.In examination of altered document,the brands of questioned ink were forecasted by classification model based on its spectral and chromatographic data,which used to determine whether it is a altered document according to the brands of questioned ink were the same or not.The main researches are as follows:1?The spectroscopy and chromatography are complementary in the classification of blue ballpoint ink.In this paper,the 270 blue ballpoint ink were conducted with spectroscopy and chromatography for more than 2000 times,and achieved Raman spectra of the ink?810?,UV-Vis spectra?540?,TLC data?540?and GC data?540?.TLC,Raman spectroscopy,and UV-Vis spectroscopy are mainly for the dye composition,and TLC has the best distinguish rate in the three methods,which divided these ink into 35 categories.Based on the classification of the TLC,some ink could be further distinguished in combination with other two methods.In addition,GC is mainly for solvent composition,so it could be more accurate for ink distinction from the point of the composition of the ink,which shows that spectroscopy and chromatography are complementary in the classification of blue ballpoint ink.Therefore,spectroscopy and chromatography should be used as far as possible with the permission of sample in practice.2 ? A new method combined spectroscopy and chromatography with pattern recognition technology was put forward to judge whether questioned document has been altered.A relationship between 25 groups of blue ballpoint pen brands and their spectral and chromatographic data was established through pattern recognition technology,and the similarity analysis,principal component analysis and pattern recognition based on K-nearest neighbor method analysis were conducted with ChempatternTMchemometrics software to distinguish blue ballpoint ink.The study found that classification model based on the brands of blue ballpoint pen ink with the principle of K-nearest neighbor method can predict the brands of unknown blue ballpoint ink,which can judge whether it is altered through the distinction of brands of questioned ink.The model based on the UV-Vis data has the highest recognition rate of 97.32% in these models,the second one is TLC-GC,80.72%,the third is TLC,78.85%,and the last one is Raman data,74.14%.With consideration of distinction efficiency,TLC-GC has the highest rate of 48.9%,and the recognition rate is well,therefore,classification model based on TLC-GC is the first choice for prediction and distinction of brands of blue ballpoint ink.In addition,other classification model should be used as far as possible with the permission of sample in practice.3 ? The multi-stage classification was proposed for the training sample in the classification model to optimize the quality of the model.Samples of the model training set?25 groups?were secondary classified according to the classification through spectroscopy and chromatography,which made distinction rate and recognition rate of the model were increased significantly,and the quality of the model was optimized at the same time.With the secondary classification,the categories of training samples for Raman spectroscopy were up to 37,and the recognition rate was increased by15.79%,UV-Vis 's were up to 41,and rate was increased by 2.5%;TLC's were up to 82,and rate was increased by 40.64%;TLC-GC's were up to 93,and rate was increased by47.19%.Through the secondary classification,the same kind of inks classified by spectroscopy and chromatography were distinguished according to brands,which raised the distinction rate,thus improving the cross validation recognition rate of the model,and optimize the quality of the model.Therefore,the multi-stage classification for training samples is an effective way to improve the quality of the model and accuracy of prediction for the brands of blue ballpoint ink through the classification model.4 ? The thought to examine the altered documents according to the prediction and distinction of brands of inks was put forward.In questioned document examination,mainly through comparing the ink formula from sample and questioned material and determine whether a questioned document has been altered.However,some different brands of blue ballpoint ink had the same type of spectrum and chromatogram through the examination of 270 blue ballpoint inks,which means the ink formula of blue ballpoint pen and their brands were not uniqueness in ourcountry,different brands of blue ballpoint pen may have the same ink formula.For altered document formed by these inks,it is difficult to judge through spectroscopy and chromatography.Therefore,the distinction of brands of blue ballpoint ink was chose as the intent and purpose in this article,with introduction of the pattern recognition in altered document examination,and through the establishment of classification model based on spectroscopy and chromatography to distinguish the questioned ink brands,which provides a new thought and method on altered document examination.In the study,the classification of different brands models were established by pattern recognition technology with data analysis and information mining for the spectral and chromatographic data of 270 different brands of blue ballpoint inks.With the prediction and distinction of brands of questioned ink by the classification model,we can make a judgment whether the questioned document has been altered.The new method proposed that combined spectroscopy and chromatography with pattern recognition technology to examine the altered document not only transferred the spectrum and chromatogram into data information,but introduced the thought of big data and artificial intelligence into altered document examination,Which can provide data support to make more objective and comprehensive judgment and provide technical support for informatization,standardization in altered document examination.