目的 解决普通拉曼光谱对牛皮纸袋进行分类的过程中存在的荧光干扰强、数据维度高、模型泛化能力不足等问题。方法 建立一种基于“差分拉曼光谱-主成分分析-深度森林”（DRS-PCA-深度森林）三级联合分类框架。结果 利用差分拉曼光谱采集54个牛皮纸袋样品的光谱数据,根据样品填料成分将其分为六大类,主成分分析将原始的1 912维光谱数据降至13维,有效降低了数据复杂度,深度森林模型在分层划分训练集和验证集条件下,其准确率达到93.2%,加权F1为0.932,显著优于SVM、随机森林等传统方法。同时,通过噪声实验验证了该模型在噪声干扰的情况下仍然能保持较高的准确率。结论 该方法为牛皮纸袋样品分类提供了高效、无损的解决方案,也可用于物证溯源及废纸回收等领域,还可为公安机关实际办案提供技术支持。
Abstract
The work aims to solve the problems of strong fluorescence interference, high data dimensionality, and insufficient model generalization ability in the classification of kraft paper bags. A three-level joint classification framework based on "DRS-PCA-Deep Forest" was built. Firstly, differential Raman spectroscopy technology was used to collect spectral data of 54 kraft paper bags samples, which were then classified into six categories according to the filler composition. Secondly, principal component analysis was used to reduce the original 1 912 dimensional spectral data to 13 dimensions, effectively reducing data complexity. Under the conditions of hierarchical division of training and validation sets, the model accuracy reached 93.2%, and the weighted F1 value was 0.932, significantly better than traditional methods such as SVM and random forest. Meanwhile, the noise experiment verified that the model could still maintain a high accuracy rate in the presence of noise interference. This method provides an efficient and non-destructive solution for kraft paper bag classification and can be extended to the judiciary. It can also be used in the fields of physical evidence tracing and waste paper recycling.
关键词
差分拉曼光谱 /
牛皮纸袋 /
主成分分析 /
深度森林模型
Key words
differential Raman spectroscopy /
kraft paper bag /
principal component analysis /
deep forest model
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 王霄, 刘蕾, 陈彭波, 等. 拉曼光谱在文件检验中的应用及进展[J]. 中国司法鉴定, 2024(4): 29-39.
WANG X, LIU L, CHEN P B, et al.Application and Research Progress of Raman Spectroscopy in the Field of Forensic Document Examination[J]. Chinese Journal of Forensic Sciences, 2024(4): 29-39.
[2] 尹宝华, 郭洪玲, 林雷祥, 等. 基于理化检验的纸张物证的比对和分类研究[J]. 刑事技术, 2017, 42(2): 124-128.
YIN B H, GUO H L, LIN L X, et al.Physicochemical Analysis to Distinguish and Classify Paper Evidence[J]. Forensic Science and Technology, 2017, 42(2): 124-128.
[3] 田国辉, 陈亚杰, 冯清茂. 拉曼光谱的发展及应用[J]. 化学工程师, 2008, 22(1): 34-36.
TIAN G H, CHEN Y J, FENG Q M.Development and Application of Raman Technology[J]. Chemical Engineer, 2008, 22(1): 34-36.
[4] 付钧泽, 姜红, 刘峰, 等. 差分拉曼光谱结合系统聚类检验香烟水松纸[J]. 化学研究与应用, 2020, 32(11): 1973-1978.
FU J Z, JIANG H, LIU F, et al.Discrimination of Cigarette Tipping Paper by Differential Raman Spectroscopy Combined with System Clustering[J]. Chemical Research and Application, 2020, 32(11): 1973-1978.
[5] XUE Q S, YU G T, LU F Q, et al.Fluorescent Labelling Combined with Confocal Differential Raman Spectroscopy to Detect Microplastics in Seawater[J]. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, 2024, 320: 124591.
[6] SOWOIDNICH K, TOWRIE M, MATOUSEK P.Shifted Excitation Raman Difference Spectroscopy Combined with Wide Area Illumination and Sample Rotation for Wood Species Classification[J]. Applied Spectroscopy, 2023, 77(6): 666-681.
[7] 郭鹏, 姜红. 拉曼光谱检验烟用内衬纸的研究[J]. 中华纸业, 2016, 37(12): 53-58.
GUO P, JIANG H.A Study on Analysis of Inner Liner for Cigarette by Raman[J]. China Pulp & Paper Industry, 2016, 37(12): 53-58.
[8] 吴泳微, 袁琨, 王坚, 等. 基于塑料近红外光谱的判别分类研究[J]. 包装工程, 2024, 45(9): 171-177.
WU Y W, YUAN K, WANG J, et al.Discriminative Classification of Plastics Based on Near-Infrared Spectra[J]. Packaging Engineering, 2024, 45(9): 171-177.
[9] 吴梦超, 屈永波, 瞿小阳, 等. 基于光谱融合的水性油墨印刷品颜色变化预测研究[J]. 包装工程, 2024, 45(17): 200-208.
WU M C, QU Y B, QU X Y, et al.Color Change Prediction of Water-Based Ink Prints Based on Spectral Fusion[J]. Packaging Engineering, 2024, 45(17): 200-208.
[10] LIEBER C A, MAHADEVAN-JANSEN A.Automated Method for Subtraction of Fluorescence from Biological Raman Spectra[J]. Applied Spectroscopy, 2003, 57(11): 1363-1367.
[11] FRANK O, JEHLIČKA J, EDWARDS H G M. Raman Spectroscopy as Tool for the Characterization of Thio-Polyaromatic Hydrocarbons in Organic Minerals[J]. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, 2007, 68(4): 1065-1069.
[12] 田陆川, 姜红, 张馨艺, 等. 差分拉曼光谱检验7种芬太尼类新精神活性物质[J]. 中国测试, 2023, 49(10): 77-82.
TIAN L C, JIANG H, ZHANG X Y, et al.Detection of Seven New Psychoactive Substances of Fentanyl by Differential Raman Spectroscopy[J]. China Measurement & Test, 2023, 49(10): 77-82.
[13] 张旭. 混合数据的主成分分析——以Gironde统计调查数据为例[D]. 广州: 暨南大学, 2016: 2-5.
ZHANG X.Principal Component Analysis of Mixed data - Taking Gironde Statistical Survey Data as an Example[D]. Guangzhou: Jinan University, 2016: 2-5.
[14] 刘广东, 邱晓晖. 基于多模式LBP与深度森林的指静脉识别[J]. 计算机技术与发展, 2018, 28(7): 83-87.
LIU G D, QIU X H.Finger Vein Recognition Based on Multi-Mode LBP and Deep Forest[J]. Computer Technology and Development, 2018, 28(7): 83-87.
[15] 朱晓妤, 严云洋, 刘以安, 等. 基于深度森林模型的火焰检测[J]. 计算机工程, 2018, 44(7): 264-270.
ZHU X Y, YAN Y Y, LIU Y A, et al.Flame Detection Based on Deep Forest Model[J]. Computer Engineering, 2018, 44(7): 264-270.
[16] 李锦, 姜红, 思沐, 等. 手持式差分拉曼光谱对纸张物证的分类研究[J]. 化学研究与应用, 2021, 33(10): 1883-1888.
LI J, JIANG H, SI M, et al.Research on Classification of Paper Evidence by Handheld Differential Raman Spectroscopy[J]. Chemical Research and Application, 2021, 33(10): 1883-1888.
[17] 张进, 姜红, 刘峰, 等. 差分喇曼光谱结合化学计量学检验烟用内衬纸[J]. 激光技术, 2021, 45(1): 61.
ZHANG J, JIANG H, LIU F, et al.Differential Raman Spectroscopy Combined with Stoichiometry for Inspection of Cigarette Liner[J]. Laser Technology, 2021, 45(1): 61.
[18] 王方原, 张敬义, 叶松, 等. 基于全光谱主成分分析的拉曼信号提取方法[J]. 光谱学与光谱分析, 2024, 44(12): 3327-3332.
WANG F Y, ZHANG J Y, YE S, et al.Raman Signal Extraction Method Based on Full Spectrum Principal Component Analysis[J]. Spectroscopy and Spectral Analysis, 2024, 44(12): 3327-3332.
[19] 王志强, 程妍昕, 张睿挺, 等. 拉曼光谱结合荧光背景对白酒品质快速检测分析[J]. 光谱学与光谱分析, 2023, 43(12): 3770-3774.
WANG Z Q, CHENG Y X, ZHANG R T, et al.Rapid Detection and Analysis of Chinese Liquor Quality by Raman Spectroscopy Combined with Fluorescence Background[J]. Spectroscopy and Spectral Analysis, 2023, 43(12): 3770-3774.
基金
食品药品安全防范山西省重点实验室开放课题（2022040709510106）
PDF(992 KB)
渝公网安备 50010702501716号 渝ICP备15012534号-2
