基于机器视觉的加热卷烟烟支端部质量检测系统设计
Design of a quality inspection system for heated cigarette ends based on machine vision
-
摘要: 针对加热卷烟烟支端部常见质量缺陷(烟支变形、空头、空松等)难以进行在线检测的问题,基于机器视觉检测技术设计了一种加热卷烟烟支端部质量检测系统。该系统利用高速计数卡、工业相机、爆闪控制器等硬件完成图像采集,在工控机中先使用Canny算法进行烟支轮廓检测识别,而后分别基于轮廓半径均值和标准差判定烟支变形、基于全局阈值二值化识别烟支空头、基于局部自适应二值化识别烟支空松,根据识别结果对烟支变形、空头、空松的端部缺陷进行在线剔除。对该系统所使用的二值化算法的性能及系统在实际生产中的应用情况进行验证,结果表明:与OTSU、Bernsen、Niblack等方法相比,全局阈值二值化在空头检测上具有最高的准确率(99.8%),自适应二值化在空松检测上具有最高的准确率(99.0%);该系统对加热卷烟烟支端部变形、空头及空松等缺陷的检测准确率≥99%,并在计算时间上有明显优势,适应于生产线高速运行要求,可为提升加热卷烟烟支端部质量及生产过程控制提供参考。Abstract: A heated cigarette end quality inspection system based on machine vision detection technology was designed to address the difficulty of online detection of common quality defects such as cigarette deformation, hollowing, and looseness at the end of heated cigarettes. The system utilized hardware such as high-speed counting cards, industrial cameras, and flash controllers to complete image acquisition, and software was set up in the industrial computer for image processing and end quality detection. Firstly, the Canny algorithm was used for cigarette contour detection and recognition in the industrial computer. Then, cigarette deformation was determined based on the mean and standard deviation of the contour radius, cigarette hollow was identified based on global threshold binarization, and cigarette looseness was identified based on local adaptive binarization. Based on the recognition results, the end defects of cigarette deformation, hollowing, and looseness were eliminated online. The performance of the binary algorithm used in the system and its practical application in production were validated. The results showed that compared with OTSU, Bernsen, Niblack and other methods, global threshold binarization had the highest accuracy (99.8%) in hollow detection, and adaptive binarization had the highest accuracy (99.0%) in looseness detection. The system had a detection accuracy of ≥99% for defects such as deformation, hollowing, and looseness of heated cigarette ends, and had significant advantages in calculation time. It was suitable for high-speed operation requirements of production lines and could provide support for improving the quality of heated cigarette ends and production process control.
-
Key words:
- heated cigarette /
- machine vision /
- cigarette end quality /
- Canny algorithm /
- binary algorithm
-
-
[1]
武凯,牟定荣,王晓辉,等.卷烟端部落丝量与卷制工艺参数的关系[J].烟草科技,2008,41(4):16-18.
-
[2]
喻赛波,王诗太,金勇,等.接装纸透气度及烟丝结构对细支卷烟逐口吸阻波动的影响[J].烟草科技,2019,52(1):79-84.
-
[3]
曹芸,张劲,王成虎,等.烟叶原料种类对烟草颗粒热解和释烟特性的影响[J].烟草科技,2022,55(4):42-50.
-
[4]
湖北中烟工业有限责任公司.一种电加热卷烟“2+2”卷制-复合-接装工艺: 202110555580.0[P].2021-08-24.
-
[5]
卢晓波,徐海,朱俊召,等.小批量生产模式下柜式先进先出卷烟储存输送装置的改进[J].烟草科技,2023,56(5):102-106.
-
[6]
张涛,张昆.一种烟支空头检测器设计[J].电子科学技术,2015,2(2):224-225.
-
[7]
任炜,冯丽辉,许胜善.烟支松头检测装置的研发与实现[J].重庆工学院学报(自然科学版),2007,21(10):97-99
,152. -
[8]
俞忠民,余其旺,熊浩.新型红外光电空头检测在PROTOS70上的应用[J].软件导刊,2011,10(2):98-100.
-
[9]
章磊,李耀,刘光徽.基于机器视觉的烟支检测系统的设计[J].电子技术应用,2012,38(5):15-18.
-
[10]
卢凡.基于机器视觉的包装机空头烟支检测技术研究[J].轻工机械,2010,28(2):65-67
,71. -
[11]
胡龙.基于机器视觉的烟支缺陷自动检测技术研究[D].长沙:湖南大学,2016.
-
[12]
曹维林,李捷,孙顺凯,等.基于Canny算子的滤棒数量检测方法[J].烟草科技,2020,53(1):96-102.
-
[13]
电子科技大学.基于深度学习的移动平台烟草激光码智能识别方法及装置:201510025849.9[P].2017-08-25.
-
[14]
邸成良,严伟,胡松,等.自适应烟丝宽度测量方法及其应用研究[J].中国激光,2014,41(7):200-204.
-
[15]
王晖,程小虎,赵淑华,等.基于机器视觉的接装纸缺陷检测装置[J].烟草科技,2015,48(8):88-92.
-
[16]
汪冬冬,侯加文,李帆,等.基于阴影检测的传送带烟丝堵料视觉检测系统设计[J].轻工学报,2020,35(5):41-47.
-
[17]
叶松涛.接装纸涂胶在线检测及剔除系统的设计应用[J].烟草科技,2012,45(3):25-27.
-
[18]
安帅帅,李庆忠.基于改进Canny的彩色图像边缘检测算法[J].软件导刊,2023,22(2):8-14.
-
[19]
卢迪,黄鑫,柳长源,等.基于区域对比度增强的二值化算法[J].电子与信息学报,2017,39(1):240-244.
-
[20]
李雷达,殷杨涛,吴金建,等.掩膜融合下的人脸图像质量评价方法[J].中国图象图形学报,2022,27(12):3476-3490.
-
[21]
王序哲.局部自适应二值化方法研究[J].软件导刊,2011,10(11):13-14.
-
[22]
庞惠文,张增红.基于数字图像处理的条码图像二值化处理研究[J].轻工科技,2021,37(6):65-66.
-
[23]
汪仕宇.基于卷积神经网络的废弃瓶分类研究与应用[D].宁波:中国科学院大学(中国科学院宁波材料技术与工程研究所),2022.
-
[24]
曹逸凡,许宝杰,徐小力,等.基于改进Bernsen算法的图像二值化研究[J].设备管理与维修,2017(18):26-28.
-
[25]
丁登峰,周国鹏,张建权,等.基于迭代思想的自适应Niblack算法改进[J].计算机应用与软件,2023,40(3):308-315.
-
[26]
李艺杰,邹坤霖,孙炜,等.基于Sauvola算法和神经网络的图像自适应二值化方法[J].测控技术,2020,39(8): 62-69
,75.
-
[1]
-
点击查看大图
计量
- PDF下载量: 5
- 文章访问数: 229
- 引证文献数: 0
下载:
