浏览全部资源
扫码关注微信
中国人民公安大学 侦查学院,北京 100038
高树辉,博士,教授,研究方向:刑事科学技术,E-mail:gaoshuhui@ppsuc.edu.cn
收稿日期:2024-09-25,
修回日期:2024-10-19,
录用日期:2024-11-08,
纸质出版日期:2025-05-15
移动端阅览
李昌盛,高树辉.基于可见光-近红外高光谱成像技术的文书朱墨时序检验[J].分析测试学报,2025,44(05):781-793.
LI Chang-sheng,GAO Shu-hui.Determining the Sequence of Intersecting Lines Based on Vis-NIR Hyperspectral Imaging Technology[J].Journal of Instrumental Analysis,2025,44(05):781-793.
李昌盛,高树辉.基于可见光-近红外高光谱成像技术的文书朱墨时序检验[J].分析测试学报,2025,44(05):781-793. DOI: 10.12452/j.fxcsxb.240925415.
LI Chang-sheng,GAO Shu-hui.Determining the Sequence of Intersecting Lines Based on Vis-NIR Hyperspectral Imaging Technology[J].Journal of Instrumental Analysis,2025,44(05):781-793. DOI: 10.12452/j.fxcsxb.240925415.
刑事文书检验领域中,文字墨迹与印章印文形成时序的分析是验证文书物证真伪的关键技术。该文基于可见光-近红外高光谱成像技术(Vis-NIR HSI)图谱合一优势,结合卷积神经网络(CNN)研究了朱墨时序的判别问题。在光谱影像形态分析的基础上,采集了42 000个不同朱墨时序样品的高光谱数据,建立朱墨时序高光谱数据集。分别使用中值滤波、Savitzky-Golay平滑滤波、多元散射校正和归一化方法对样本光谱进行预处理;采用连续投影算法(SPA)和竞争自适应重加权采样(CARS)对光谱进行特征波长选择,分别建立逻辑回归(LR)等若干二分类机器学习模型和一维卷积神经网络(1D-CNN)模型,并比较了建模效果。实验结果显示,基于CARS方法提取的光谱特征波长建立的CARS-1D-CNN模型在训练集和测试集上的准确率分别达96.98%和95.54%,表明Vis-NIR HSI与1D-CNN结合能够有效识别朱墨时序。该方法与常规检验方法相互辅助、相互验证,能够提高朱墨时序检验鉴定的准确性和效率。
In the field of forensic questioned document examination,the analysis of the ink of the text and the seal imprint formation time sequence is the key technology to verify the authenticity of the instrument material evidence. This paper proposes a sequence of intersecting lines identification model that combines visible-near infrared hyperspectral imaging(Vis-NIR HSI) and convolutional neural networks(CNN). A dataset of hyperspectral data from 42 000 different red-black ink sequence samples was collected.The samples were preprocessed using median filtering(MF),Savitzky-Golay smoothing(SG smoothing),multiplicative scatter correction(MSC),and normalization methods. Further,the successive projection algorithm(SPA) and competitive adaptive reweighted sampling(CARS) were employed for spectral feature wavelength selection. Several binary classification machine learning models,including logistic regression(LR),and a one-dimensional convolutional neural network(1D-CNN) model were developed and their modeling effects compared. The results showed that the CARS-1D-CNN model,established using feature wavelengths extracted by the CARS method,achieved the best classification performance with accuracies of 96.98% on the training set and 95.54% on the test set. The study demonstrates that the combination of Vis-NIR HSI and 1D-CNN can effectively identify the sequence of intersecting lines. This method,in conjunction with conventional examination techniques,can further enhance the accuracy and efficiency of forensic questioned document examination.
Xie P , Hu Z P . The Identification of the Age of the Questioned Document . Hangzhou : Zhejiang University Press (谢朋,胡祖平. 文件制成时间检验. 杭州:浙江大学出版社), 2014 : 7 .
Tian J , Su H T , You J D , Cao S R . Guangdong Gongan Keji (田洁,苏泓滔,游嘉德,曹淑瑞. 广东公安科技), 2022 , 30 ( 4 ): 8 - 12 .
Song C Y , Bai X F . Chin. J. Forensic Sci. (宋朝阳 , 白晓峰 . 中国司法鉴定) , 2018 ,( 3 ): 56 - 60 .
Wu X , Fang F , Li B . J. Forensic Sci. , 2019 , 64 ( 6 ): 1761 - 1768 .
Li B . J. Forensic Sci. , 2016 , 61 ( 3 ): 809 - 814 .
Han W , Huang J T , Chen W N . Phys. Test. Chem. Anal.:Chem. Anal. (韩伟,黄建同,陈维娜. 理化检验-化学分册), 2015 , 51 ( 12 ): 1753 - 1759 .
Wang Y , Li B . Sci. Justice , 2012 , 52 ( 2 ): 112 - 118 .
Jiang P , Wang W X , Zhang N , Wang C M , Huang W , Shi G J , Xu X J , Gao S H . Laser Optoelectron . Prog. (蒋鹏,王蔚昕,张宁,王成铭,黄威,石高军,许小京,高树辉. 激光与光电子学进展), 2020 , 57 ( 22 ): 260 - 268 .
Tao K M , Yan Y H . Forensic Sci. Technol. (陶克明 , 阎育华 . 刑事技术) , 1989 ,( 4 ): 13 - 14 .
Fang J , Xu Y M . Forensic Sci . Technol. (方健 ,许耀明. 刑事技术), 1994 ,( 4 ): 27 .
Ouyang G L . J. Henan Judicial Police Vocat . Coll . (欧阳国亮. 河南司法警官职业学院学报), 2020 , 18 ( 2 ): 102 - 106 .
Li B , Feng M S . Guangdong Gongan Keji (李彪,冯明帅.广东公安科技), 2012 , 20 ( 1 ): 10 - 17 .
Sun H Q , Dong J , Xu Y Y , Shao S . China Public Secur. : Acad . Ed. (孙华清,董军,许燕燕,邵帅. 中国公共安全 : 学术版) , 2012 ,( 1 ): 113 - 116 .
Xie P , Li B . J . People's Public Secur . Univ . China : Sci. Technol. (谢朋,李彪. 中国人民公安大学学报:自然科学版), 2003 ,( 2 ): 36 - 37 .
Mariotti K D C , Ortiz R S , Ferrão M F . Sci. Justice , 2023 , 63 ( 3 ): 387 - 395 .
Liu K K , Luo Y P . Laser Optoelectron . Prog. (刘康康,罗亚平. 激光与光电子学进展), 2024 , 61 ( 4 ): 58 - 65 .
Zhang H , Gao S H . J. Instrum. Anal. (张浩,高树辉. 分析测试学报), 2023 , 42 ( 7 ): 817 - 824 .
Zhuang Y , Gao S H , Xie F , Huang W , Cai N B . Laser Optoelectron . Prog. (庄园,高树辉,谢菲,黄威,蔡能斌. 激光与光电子学进展), 2022 , 59 ( 16 ): 484 - 492 .
Melit D B , George S . Forensic Sci. Int. , 2020 , 311 : 110194 .
Wang S , He H , Lv R , He W , Li C , Cai N . J. Forensic Sci. , 2022 , 67 ( 2 ): 550 - 561 .
Yaseen M , Aftab R , Mahrukh R . arXiv , 2006 . https://arxiv.org/abs/2006.16057 https://arxiv.org/abs/2006.16057 .
Zhao F , Yan Z R , Song H Y . Spectrosc . Spectral Anal. (赵凡,闫昭如,宋海燕.光谱学与光谱分析), 2021 , 41 ( 7 ): 2240 - 2244 .
Chen R J , Pu Y W , Wu F Z , Liu Y C , Li Q . Laser . Optoelectron. Prog. (陈如俊,普运伟,吴锋振,刘昱岑,李奇.激光与光电子学进展), 2023 , 60 ( 16 ): 164 - 172 .
Lecun Y , Bottou L , Bengio Y , Haffner P . Proc. IEEE , 1998 , 86 ( 11 ): 2278 - 2324 .
Huang J , He H , Lv R , Zhang G , Zhou Z , Wang X . Anal. Chim. Acta , 2022 , 1224 : 340238 .
Mao A , Mohri M , Zhong Y . arXiv , 2023 . http://arxiv.org/abs/2304.07288 http://arxiv.org/abs/2304.07288 .
Kingma D P , Ba J . arXiv , 2017 . http://arxiv.org/abs/1412.6980 http://arxiv.org/abs/1412.6980 .
Hu X D , Yin L , Yang Q C , Wang L . Laser Optoelectron . Prog. (胡锡敦,尹禄,杨钦晨,王乐. 激光与光电子学进展), 2024 , 61 ( 2 ): 460 - 472 .
Chu X L . Chemometric Methods in Modern Spectral Analysis . Beijing : Chemical Industry Press (褚小立. 现代光谱分析中的化学计量学方法. 北京:化学工业出版社), 2022 : 121 , 124 .
0
浏览量
12
下载量
0
CSCD
关联资源
相关文章
相关作者
相关机构