目的 解决高速卷烟包装生产中烟包铝箔纸因高反光、柔性特质导致的褶皱、破损等外观缺陷在线检测难题,以及传统视觉检测方法误检率高、深度学习算法在复杂工况下泛化性不足的现状。方法 提出一种融合线激光投影统计学与YOLOv7目标检测算法的双光源视觉检测方法。该方法通过构建双光源成像系统,基于三角测量原理将表面微观高度变化转换为激光线像素位移;继而利用投影统计特征（标准差、梯度等）实现缺陷快速初筛;最后引入YOLOv7算法,利用其扩展聚合网络（ELAN）与重参数化卷积（RepConv）结构,对复杂及统计学特征不明显的缺陷进行精准识别,形成统计学初筛与深度学习精判的双重检测机制。结果 实验结果表明,该系统在400包/min的生产速度下稳定运行,铝箔纸缺陷检出率超过99.9%,下游工序缺陷返回数量由安装前的日均1.6包降为0。结论 本研究有效解决了铝箔纸高反光干扰下的缺陷检出问题,为烟草包装高反光材质表面缺陷在线检测提供了高精度、高鲁棒性的可靠方案。
Abstract
The work aims to deal with the challenges of online inspection for wrinkles, breaks, and other surface defects on cigarette package aluminum foil arising from its highly reflective and flexible nature, as well as the high false detection rates of traditional visual inspection methods and the insufficient generalization of deep learning algorithms in complex working conditions. A dual-light source visual inspection method that integrated line laser projection statistics with the YOLOv7 object detection algorithm was proposed. Firstly, an imaging system consisting of a white light source and a red line laser was constructed. Based on the triangulation principle, microscopic height variations on the foil surface were converted into pixel displacements of the laser line. Subsequently, by extracting projection statistical features (e.g., standard deviation, gradient) of the laser line centerline, a rapid defect classification rule was established to achieve preliminary screening of typical defects such as wrinkles and breaks. Finally, the YOLOv7 deep learning model was introduced, leveraging its Extended Efficient Layer Aggregation Network (ELAN) and Reparameterized Convolution (RepConv) structures to accurately locate and identify complex defects, forming a dual detection mechanism of statistical preliminary screening and deep learning refined judgment. Industrial experimental results demonstrated that the system operated stably at a production speed of 400 packs per minute, achieving a defect detection rate exceeding 99.9%. The number of defective packs returned from downstream processes was reduced from a daily average of 1.6 before installation to zero, while the false removal rate remained very low. This research effectively addresses the problem of defect detection under strong interference from aluminum foil reflections, providing a reliable, high-precision, and highly robust solution for online quality control in tobacco packaging.
关键词
铝箔纸缺陷 /
线激光投影 /
投影统计学 /
YOLOv7 /
在线检测 /
机器视觉
Key words
aluminum foil defects /
line laser projection /
projection statistics /
YOLOv7 /
online inspection /
machine vision
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 刘俊华, 李东升, 王建华. 机器视觉技术在工业检测中的应用与发展综述[J]. 仪器仪表学报, 2020, 41(8): 1-16.
LIU J H, LI D S, WANG J H.Review on the Application and Development of Machine Vision Technology in Industrial Inspection[J]. Chinese Journal of Scientific Instrument, 2020, 41(8): 1-16.
[2] 叶家俊. 光度立体成像技术实现高效精准2.5D缺陷检测[J/OL]. 激光杂志, 2025: 1-8[2025-12-16]. https://kns.cnki.net/kcms2/article/abstract v=3uoqIhG8 C44YLTlOAiTRKibYlV5Vjs7i8oJR1HYAr80I5mqSfqL7wqObWLrPZq1L0lQw7S8IxP7Lk8A0p9Y1JQ.
YE J J. Photometric Stereoscopic Imaging Technology for Efficient and Accurate 2.5D Defect Detection[J/OL]. Laser Journal, 2025: 1-8[2025-12-16]. https://kns.cnki.net/ kcms2/article/abstract v=3uoqIhG8C44YLTlOAiTRKib YlV5Vjs7i8oJR1HYAr80I5mqSfqL7wqObWLrPZq1L0lQw7S8IxP7Lk8A0p9Y1JQ.
[3] 陈真诚, 倪旭翔, 陆祖康. 基于线结构光视觉传感器的表面缺陷检测系统[J]. 光学学报, 2006, 26(1): 63-67.
CHEN Z C, NI X X, LU Z K.Surface Defect Inspection System Based on Line Structured Light Vision Sensor[J]. Acta Optica Sinica, 2006, 26(1): 63-67.
[4] WANG C Y, BOCHKOVSKIY A, LIAO H Y M. YOLOv7: Trainable Bag-of-Freebies Sets New State-of-the-Art for Real-Time Object Detectors[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2022: 7464-7475.
[5] 全国有色金属标准化技术委员会(SAC/TC 243). 铝箔试验方法第2部分:针孔的检测: GB/T 22638.2—2016[S]. 北京: 质检出版社, 2016.
National Nonferrous Metals Standardization Technical Committee. Test methods for aluminium foil—Part 2: Determination of pinholes: GB/T 22638.2-2016[S]. Beijing: Quality Inspection Press, 2016.
[6] 全国有色金属标准化技术委员会. 烟包装用铝箔针孔度检测的重要性与测试仪应用: YS/T 846—2025[S]. 北京: 中国标准出版社, 2025.
National Nonferrous Metals Standardization Technical Committee. The Importance of Pinhole Detection of Aluminum Foil for Cigarette Packaging and the Application of the Tester: YS/T 846-2025[S]. Beijing: Standards Press of China, 2025.
[7] 毛文杰, 谢世龙, 李林彧璇, 等. 基于深度学习的陶瓷小目标缺陷检测算法及实验[J]. 科学技术与工程, 2025, 25(11): 4666-4672.
MAO W J, XIE S L, LI L Y X, et al. Ceramic Small-Target Defect Detection Algorithm and Experiment Based on Deep Learning[J]. Science Technology and Engineering, 2025, 25(11): 4666-4672.
[8] 魏烨, 秦少谦, 吕盼稂. 基于深度学习的3D激光测量图像反光区域分离方法[J]. 工业控制计算机, 2020, 33(9): 47-50.
WEI Y, QIN S Q, LYU P L.Reflective region separation method for 3D laser measurement images based on deep learning[J]. Industrial Control Computer, 2020, 33(9): 47-50.
[9] 胡寅, 王永红, 刘济林. 基于三角测量的线激光传感器标定方法[J]. 浙江大学学报(工学版), 2016, 50(1): 19-25.
HU Y, WANG Y H, LIU J L.Calibration Method of Line Laser Sensor Based on Triangulation[J]. Journal of Zhejiang University (Engineering Science), 2016, 50(1): 19-25.
[10] 孙晓亮, 韩震, 刘巍等. 基于线结构光视觉的钢板表面缺陷在线检测系统[J]. 仪器仪表学报, 2019, 40(3): 90-99.
SUN X L, HAN Z, LIU W, et al.On-line Detection System for Steel Plate Surface Defects Based on Line Structured Light Vision[J]. Chinese Journal of Scientific Instrument, 2019, 40(3): 90-99.
[11] 张明, 王磊, 刘洋. 基于线结构光的金属表面微小缺陷高精度检测方法[J]. 仪器仪表学报, 2023, 44(2): 88-95.
ZHANG M, WANG L, LIU Y.A High-Precision Detection Method for Micro-Defects on Metal Surfaces Based on line Structured Light[J]. Chinese Journal of Scientific Instrument, 2023, 44(2): 88-95.
[12] 王健, 李芳, 刘强. 基于蓝色线激光的烟包卡纸表面缺陷检测系统[J]. 光学精密工程, 2022, 30(5): 123-132.
WANG J, LI F, LIU Q.Surface Defect Detection System for Cigarette Pack Cardstock Based on Blue Line Laser[J]. Optics and Precision Engineering, 2022, 30(5): 123-132.
[13] 李伟, 张磊. 基于机器视觉的铝箔纸撕裂缺陷检测算法[J]. 计算机工程与应用, 2023, 59(10): 234-240.
LI W, ZHANG L.Detection Algorithm for Tear Defects of Aluminum Foil Paper Based on Machine Vision[J]. Computer Engineering and Applications, 2023, 59(10): 234-240.
[14] REDMON J, DIVVALA S, GIRSHICK R, et al.You Only Look Once: Unified, Real-Time Object Detection[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016: 779-788.
[15] 陆海华, 黄春辉, 王旭东, 等. 基于特征增强与多尺度融合的烟包外观缺陷检测方法研究[J]. 中国烟草学报, 2025, 31(4): 41-50.
LU H H, HUANG C H, WANG X D, et al.Research on Cigarette Pack Appearance Defect Detection Method Based on Feature Enhancement and Multi-Scale Fusion[J]. Acta Tabacaria Sinica, 2025, 31(4): 41-50.
[16] 宗国浩, 张明琰, 王锐, 等. 卷烟包装外观缺陷数据集构建及深度学习检测技术研究[J]. 包装工程, 2024, 45(5): 135-143.
ZONG G H, ZHANG M Y, WANG R, et al.Construction of a Cigarette Packaging Appearance Defects Dataset and Research on Deep Learning Detection Technology[J]. Packaging Engineering, 2024, 45(5): 135-143.
[17] 刘泽, 左雯, 李思源, 等. 一种摊薄机构中烟梗摊薄率的表征方法及用途: 中国, CN115868659B[P].2025-05-13.
LIU Z, ZUO W, LI S Y, et al. A Characterization Method and Application of Tobacco Stem Thinning Rate in a Thinning Mechanism: China, CN115868659B[P].2025-05-13.
[18] 杨艳, 刘守阳, 张旭, 等. 基于YOLOv7改进的PCB表面缺陷实时检测方法[J]. 电子测量与仪器学报, 2023, 37(5): 1-10.
YANG Y, LIU S Y, ZHANG X et al. Real-Time Detection Method for PCB Surface Defects Based on Improved YOLOv7[J]. Journal of Electronic Measurement and Instrumentation, 2023, 37(5): 1-10.
[19] 中烟物流技术有限责任公司, 重庆烟叶复烤有限公司. “基于多模态数据融合算法片烟装箱密度检测控制研究”项目启动会召开[EB/OL]. (2025-01-10) [2025-12-16]. http://www.eastobacco.com/kjxw/202501/t20250110_556789.html.
China Tobacco Logistics Technology Co., Ltd., Chongqing Tobacco Leaf Re-drying Co., Ltd. Kick-off meeting for the project "Research on Density Detection and Control of Cigarette Packaging Based on Multimodal Data Fusion Algorithm" was Held[EB/OL]. (2025-01-10)[2025-12-16]. http://www.eastobacco.com/ kjxw/202501/t20250110_556789.html.
[20] 何凤梅, 陈逸怀. 基于激光条纹和深度学习的钢构件表面缺陷检测方法研究[J]. 装备制造技术, 2025(6): 5-8.
HE F M, CHEN Y H.Research on Surface Defect Detection Method of Steel Components Based on Laser Fringe and Deep Learning[J]. Equipment Manufacturing Technology, 2025(6): 5-8.
基金
红云红河烟办[2017]174号重点项目（HYHH2025GY03）
PDF(5109 KB)
渝公网安备 50010702501716号 渝ICP备15012534号-2
