目的 为克服包装废弃物材质光谱相近及表面状态干扰导致的分类精度难题,本研究采用高光谱成像技术研究不同分类模型对典型包装材质的判别能力并突破单一材质局限,开展跨材质、跨形状验证。方法 针对纸材、布材、塑料和木材四类不同材质和形状的典型包装废弃物,采用高光谱成像技术收集图谱信息,对比最大似然法（Maximum Likelihood Classification,MLC）、反向传播神经网络法（Back Propagation Neural Network,BPNN）、支持向量机（Support Vector Machine,SVM）三种分类模型的分类效果并进行分类预测。结果 第一主成分所含原始信息最丰富,前两个主成分累计贡献率达到99.79%。三种分类模型均可实现包装废弃物的分类,其中SVM的分类效果最优。采用Majority/Minority分析后处理优化,其总体分类精度高达99.91%,Kappa系数为0.998 6。经跨形状样本验证,模型泛化性强,能适配不规则废弃物的分选目标。结论 本研究证实高光谱成像技术结合机器学习模型可精准识别包装废弃物,将为利用现代科技实现精准化、智能化垃圾分类提供技术参考。
Abstract
To overcome the challenge of low classification accuracy caused by similar spectral characteristics of packaging waste materials and interference from varying surface conditions, the study employs hyperspectral imaging technology to investigate the discriminative performance and generalization stability of different classification models for typical packaging materials, breaking the limitations of single-material studies and conducting cross-material and cross-shape validation. Hyperspectral imaging technology was applied to collect spectral image data from four types of typical packaging waste with different shapes, namely paper, fabric, plastic and wood. The classification performance of three models, including Maximum Likelihood Classification (MLC), Back Propagation Neural Network (BPNN), and Support Vector Machine (SVM), was compared, and classification prediction was conducted accordingly. The first principal component contained the most of original information, and the cumulative contribution of the first two principal components reached 99.79%. All three classification models achieved valid classification of packaging waste based on the dimensionality-reduced features, among which SVM achieved the optimal performance. After post-processing optimization through Majority/Minority analysis, the overall classification accuracy reached 99.91% with a Kappa coefficient of 0.998 6. Validation with cross-shape samples demonstrated strong model generalization, suitable for sorting irregular waste. This work verifies that the hyperspectral imaging technology combined with machine learning models enables accurate identification of packaging waste, providing a technical reference for the precise and intelligent sorting of solid waste with modern technologies.
关键词
高光谱成像技术 /
包装废弃物 /
分类模型 /
主成分分析
Key words
hyperspectral imaging technology /
packaging waste /
classification model /
principal component analysis
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 熊志文. 我国包装废弃物资源化利用管理政策研究[D]. 赣州: 江西理工大学, 2011.
XIONG Z W.Research on the Management Policies for the Resource Utilization of Packaging Waste in China [D]. Ganzhou: Jiangxi University of Science and Technology, 2011.
[2] UNEP. Global Waste Management Outlook 2024: Beyond an age of waste - Turning rubbish into a resource[R]. Nairobi: United Nations Environment Programme, 2024. DOI: 10.59117/20.500.11822/44939.
[3] KOINIG G, KUHN N, FINK D, et al.Near-infrared Spectroscopy for Online Classification of Plastic Packaging Films[J]. Waste Management, 2022(150): 138-147.
[4] XU Z Y.Efficient and Anti-interference Plastic Classification Method Suitable for One-shot Learning Based on Laser Induced Breakdown Spectroscopy[J]. Chemosphere, 2025(378): 144412.
[5] GUNASEELAN N, PREMALATHA S, UMA P.Classification of garbage using modified ResNeXt model[J]. Journal of Environmental Management, 2022(321): 115997.
[6] 吴瑞. 基于高光谱成像和深度学习的可回收垃圾分类方法研究[D]. 太原: 中北大学, 2020.
WU R.Research on Recyclable Waste Classification Method Based on Hyperspectral Imaging and Deep Learning[D]. Taiyuan: North University of China, 2020.
[7] 张保华, 李江波, 樊书祥, 等. 高光谱成像技术在果蔬品质与安全无损检测中的原理及应用[J]. 光谱学与光谱分析, 2014(10): 2743-2751.
ZHANG B H, LI J B, FAN S X, et al.Principles and Applications of Hyperspectral Imaging Technology in Non-destructive Testing of Fruit and Vegetable Quality and Safety[J]. Spectroscopy and Spectral Analysis, 2014(10): 2743-2751.
[8] LANDGREBE D.Hyperspectral image data analysis[J]. IEEE Signal Processing Magazine, 2002, 19(1): 17-28.
[9] WEI L, WANG Z, HUANG C, et al.Transparency Estimation of Narrow Rivers by UAV-borne Hyperspectral Remote Sensing Imagery[J]. IEEE Access, 2020(8): 168137-168153.
[10] WANG L, PU H, SUN D W.Estimation of Chlorophyll-a Concentration of Different Seasons in Outdoor Ponds Using Hyperspectral Imaging[J]. Talanta, 2016(147): 422-429.
[11] LIEW S C, CHOO C K, LAU J W M, et al. Monitoring Water Quality in Singapore Reservoirs with Hyperspectral Remote Sensing Technology[J]. Water Practice & Technology, 2019, 14(1): 118-125.
[12] NAGESWARAN K, NAGARAJAN K, BANDIYA R.A Novel Algorithm for Hyperspectral Image Denoising in Medical Application[J]. Journal of Medical Systems, 2019, 43(9): 291.
[13] WEI X, LI W, ZHANG M, et al.Medical hyperspectral image classification based on end-to-end fusion deep neural network[J]. IEEE Transactions on Instrumentation and Measurement, 2019, 68(11): 4481-4492.
[14] Tamin O, Moufti A, El Mabrouk K, et al.Hyperspectral imaging for plastic waste and microplastic detection: A review[J]. Environmental Science and Pollution Research, 2023, 30(45): 100243-100261.
[15] Bonifazi G, D‘Adamo I, Palmieri R, et al. Recycling-oriented characterization of space waste through clean hyperspectral imaging technology in a circular economy context[J]. Clean Technologies, 2025, 7(1): 26.
[16] Larsen C S, et al.Long-wave infrared hyperspectral imaging for black plastic waste sorting[J]. Waste Management, 2024, 174: 1-10.
[17] Zheng Y, Bai J, Xu J, et al.A discrimination model in waste plastics sorting using NIR hyperspectral imaging system[J]. Waste Management, 2018, 72: 87-98.
[18] Mehrubeoglu M, Van Sickle A, Turner J.Detection and identification of plastics using SWIR hyperspectral imaging[C]//Imaging Spectrometry XXIV: Applications, Sensors, and Processing. SPIE, 2020, 11504: 85-95.
[19] 彭斌彬, 张潮, 郭亚坤, 等. 基于近红外光谱的废旧塑料材质识别研究[J]. 激光杂志, 2025, (2): 210-217.
PENG B B, ZHANG C, GUO Y K, et al.Study on Identification of Waste Plastic Materials Based on Near-Infrared Spectroscopy[J]. Laser Magazine, 2025, (2): 210-217.
[20] Diruit W, Burel T, Bajjouk T, et al.Comparison of supervised classifications to discriminate seaweed-dominated habitats through hyperspectral imaging data[J]. Journal of Applied Phycology, 2024.
[21] 杨洋, 徐熙平, 薛航, 等. 基于SPA-PSO-BP的花生高光谱图像分类方法研究[J]. 激光技术, 2024, 48(4): 556-564.
YANG Y, XU X, XUE H, et al.Research on Peanut Hyperspectral Image Classification Methodbased on SPA-PSO-BP[J]. Laser Technology, 2024, 48(4): 556-564.
[1] BLANZIERI E, MELGANI F.Nearest Neighbor Classification of Remote Sensing Images with the Maximal Margin Principle[J]. IEEE Transactions on Geoscience and Remote Sensing, 2008, 46(6): 1804-1811.
[22] MELGANI F, BRUZZONE L.Classification of Hyperspectral Remote Sensing Images with Support Vector Machines[J]. IEEE Transactions on Geoscience and Remote Sensing, 2004, 42(8): 1778-1790.
[23] 赵冬娥, 吴瑞, 赵宝国, 等. 高光谱成像的垃圾分类识别研究[J]. 光谱学与光谱分析, 2019, 39(3): 917.
ZHAO D E, WU R, ZHAO B G, et al.Research on Garbage Classification and Recognition Using Hyperspectral Imaging[J]. Spectroscopy and Spectral Analysis, 2019, 39(3): 917.
[24] ALIEVA F.Prediction of Plastic Fragments in Recycled Paper Using Near-Infrared Spectroscopy[D]. Karlstad: Karlstads Universitet, 2023.
[25] XU Y, XUAN C, YING Z, et al.Research on the Method of Near-Infrared Hyperspectral Classification of Cotton-Polyester Blended Waste Fabric Based on Deep Learning[J]. Recycling, 2026, 11(2): 42.
[26] GORLA G, NIELSEN F, MONTAGUE P B, et al.Challenges and Spectra Interpretability in Textile Sorting: NIR Hyperspectral Images and Chemometrics[J]. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, 2026(344): 126665.
[27] 胡锡敦, 尹禄, 杨钦晨, 等. 基于近红外高光谱成像技术的塑料分类(特邀)[J]. 激光与光电子学进展, 2024, 61(2): 0211031.
HU X D, YIN L, YANG Q C, et al.Plastic Classification Based on Near-Infrared Hyperspectral Imaging Technology (Invited)[J]. Progress in Laser and Optoelectronics, 2024, 61(2): 0211031.
[28] YE J, ZHOU Y, WANG M, et al.Advances in Multi-source Perception Technology Integrated with Machine Learning in Food Quality Evaluation[J]. Science and Technology of Food Industry, 2025, 46(23): 466-475.
[29] RICHARDS J A.Remote Sensing Digital Image Analysis: An Introduction[M]. 5th ed. Berlin: Springer- Verlag, 2013: 365-370.
基金
天津市应用基础与前沿技术研究计划重点项目（14JCZDJC34600）; 大学生创新创业训练计划项目（202510069244）
PDF(23445 KB)
渝公网安备 50010702501716号 渝ICP备15012534号-2
