目的 建立一种快速、无损的可区分可降解塑料袋与不可降解塑料袋的分类方法。方法 利用高光谱成像系统（900⁓2 500 nm）采集56个塑料袋样本的高光谱数据,采用监督自编码器将392个波段降维至32维潜变量特征;构建三级分层分类策略,分别训练多层感知器（MLP）、卷积神经网络（CNN）、循环神经网络（RNN）及注意力融合模型进行分类评估。结果 注意力融合模型在第1次分类（生物基与石油基）中准确率达96.10%,第2次分类（生物基细分）中F1值达0.962,第3次分类（石油基细分）中所有模型均达100%识别率。结论 该架构结合高光谱技术与深度集成模型,实现了可降解塑料袋的高精度自动分类,不仅可为功能性生物质基包装材料的回收与智能分选提供技术支撑,助力绿色包装废弃物资源化利用;同时也为公安物证检验提供了有效技术方案。
Abstract
The work aims to establish a rapid, non-destructive method for distinguishing between biodegradable and non-biodegradable plastic bags. Hyperspectral data of 56 plastic bag samples were acquired with a hyperspectral imaging system (900-2 500 nm). A supervised autoencoder was employed to reduce the 392 spectral bands to 32-dimensional latent variables. A three-level hierarchical classification strategy was constructed, and a multilayer perceptron (MLP), a convolutional neural network (CNN), a recurrent neural network (RNN) and an attention fusion model were trained and evaluated for classification. Results showed that in the first classification (bio-based vs. petroleum-based), the attention fusion model achieved an accuracy of 96.10%; in the second classification (bio-based sub-classification), its F1 score reached 0.962; in the third classification (petroleum-based sub-classification), all models achieved 100% recognition rate. In conclusion, the proposed architecture, combining hyperspectral technology with deep ensemble models, enables high-precision automatic classification of degradable plastic bags. It not only provides technical support for the recycling and intelligent sorting of functional biomass-based packaging materials, facilitating the resource utilization of green packaging waste, but also offers an effective technical solution for forensic evidence examination.
关键词
塑料购物袋 /
高光谱成像 /
可降解塑料 /
注意力融合模型 /
绿色包装回收 /
物证分类
Key words
plastic shopping bags /
hyperspectral imaging /
degradable plastics /
attention fusion model /
green packaging recycling /
forensic evidence classification
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 管艺博, 杜欣芸, 张志霞. 基于近红外光谱结合MS-Mamba模型的塑料包装分类方法[J]. 包装工程, 2026, 47(3):198-209.
GUAN Y B, DU X Y, ZHANG Z X.Plastic Packaging Classification Based on Near-Infrared Spectroscopy and MS-Mamba Model[J]. Packaging Engineering, 2026, 47(3):198-209.
[2] 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.
[3] 杨棋驭, 胡晓光, 姜红, 等. 基于双曲面弯晶技术的X射线荧光光谱仪对塑料快递袋检验分类[J]. 包装工程, 2024, 45(13):247-252.
YANG Q Y, HU X G, JIANG H, et al.Classification of Plastic Express Bags by X-Ray Fluorescence Spectrometer Based on Johansson-Type Doubly Curved Crystal[J]. Packaging Engineering, 2024, 45(13):247-252.
[4] 陈壮, 姜红, 罗鸿斌, 等. X射线荧光光谱法结合HCA-PCA-BPNN实现塑料快递包装袋识别分类[J]. 塑料工业, 2022, 50(11):138-144.
CHEN Z, JIANG H, LUO H B, et al.Identification and Classification of Plastic Express Bags by X-Ray Fluorescence Spectrometry Combined with HCA-PCA-BPNN[J]. China Plastics Industry, 2022, 50(11):138-144.
[5] 姜红, 马星煜, 周贯旭. 差分拉曼光谱技术在理化检验中的应用[J/OL]. 湖南警察学院学报, 2026-04-09. https://link.cnki.net/urlid/43.1511.D.20260407.1824.002.
JIANG H, MA X Y, ZHOU G X. Application of Differential Raman Spectroscopy in Physicochemical Analysis[J/OL]. Journal of Hunan Police Academy, 2026-04-09. https://link.cnki.net/urlid/43.1511.D.20260407.1824.002.
[6] 丁建明, 王鑫, 张容玲, 等. 拉曼光谱结合偏最小二乘法定量分析水中聚乙烯和聚苯乙烯两种微塑料[J]. 分析化学, 2024, 52(10):1581-1590.
DING J M, WANG X, ZHANG R L, et al.Raman Spectroscopy Combined with Partial Least Squares for Quantitative Analysis of Two Kinds of Microplastics in Water Samples[J]. Chinese Journal of Analytical Chemistry, 2024, 52(10):1581-1590.
[7] IEONG N W C, DUSTIN M, COULSON S. Analysis of Forensic Polymer Samples Using Double Shot Pyrolysis Gas Chromatography-Mass Spectrometry[J]. Forensic Science International, 2024, 364:112238.
[8] LOGRADO L P L, BRAGA J W B. Evaluation of Interferents in Sampling Materials for Analysis of Post-Explosion Residues (Explosive Emulsion/ANFO) Using Gas Chromatography-Mass Spectrometry (GC/MS)[J]. Journal of Forensic Sciences, 2025, 70(1):314-322.
[9] 陈浩洋, 高树辉. 法庭科学中光谱成像技术的应用与展望:聚焦高光谱、拉曼与太赫兹[J]. 分析测试学报, 2026, 45(4):900-916.
CHEN H Y, GAO S H.Applications and Prospects of Spectral Imaging Technology in Forensic Science:Focus on Hyperspectral, Raman, and Terahertz Imaging[J]. Journal of Instrumental Analysis, 2026, 45(4):900-916.
[10] 刘英霞, 莫燕, 贾硕果. 高光谱结合化学计量学在文件物质材料分析中的研究进展[J]. 光散射学报, 2025, 37(4):560-570.
LIU Y X, MO Y, JIA S G.Research Progress of Hyperspectral Combined with Chemometrics in the Analysis of Document Materials[J]. The Journal of Light Scattering, 2025, 37(4):560-570.
[11] 王阔, 李开开. 基于高光谱技术的不同介质血迹分类研究[J]. 激光与光电子学进展, 2026, 63(2):0230002.
WANG K, LI K K.Classification of Bloodstains in Different Media Based on Hyperspectral Technology[J]. Laser & Optoelectronics Progress, 2026, 63(2):0230002.
[12] 杨泽会, 吴箭, 李瑞东, 等. 卷积桥接孪生自编码器的近红外光谱转移研究[J]. 分析测试学报, 2025, 44(3):471-478.
YANG Z H, WU J, LI R D, et al.Study on near Infrared Spectrum Transfer of Convolutional Bridged Twin Denoising Reduction Autoencoder[J]. Journal of Instrumental Analysis, 2025, 44(3):471-478.
[13] 赵森林. 基于近红外光谱的塑料材质识别技术研究[D]. 合肥:合肥大学, 2025.
ZHAO S L.Research on Plastic Material Identification Technology Based on Near-Infrared Spectroscopy[D]. Hefei:Hefei University, 2025.
[14] 王艳霞. 新型添加型光敏剂芳基三唑制备光降解塑料及性能研究[J]. 塑料科技, 2021, 49(2):48-50.
WANG Y X.Study on Properties of Photodegradable Plastic Prepared by LDPE Doped with New Photosensitizer Aryltriazole[J]. Plastics Science and Technology, 2021, 49(2):48-50.
[15] 周飞翔, 姜红, 钟方昊, 等. 基于高光谱结合机器学习对塑料瓶盖的快速分类研究[J]. 上海塑料, 2024, 52(2):54-59.
ZHOU F X, JIANG H, ZHONG F H, et al.Fast Identification of Plastic Bottle Caps Based on Hyperspectral Combined with Machine Learning[J]. Shanghai Plastics, 2024, 52(2):54-59.
PDF(2363 KB)
渝公网安备 50010702501716号 渝ICP备15012534号-2
