摘 要:对于家具板件自动化生产,快速、准确的板件质量检测系统是不可或缺的。以 YOLO v5 算法为基础对计算机辅助检查板件质量技术进行了研究。在机器学习算法的基础上,结合运动目标捕捉算法作为辅助检测方法,实现板件的识别和定位,然后通过 OpenCV 中的图像处理和轮廓识别方法来实现板件尺寸测量。该算法的板件识别率达 98.33%,识别多种颜色板件外形长宽尺寸时,实测最大尺寸误差 2.23 mm,能满足家具板件生产质量检测的需求。
关键词:机器视觉;YOLO v5;边缘检测;尺寸检测
DOI:10.19850/j.cnki.2096-4706.2021.09.039
基金项目:广东省科技计划项目(2018A050 506024)
中图分类号:TP277 文献标识码:A 文章编号:2096-4706(2021)09-0149-05
Research on Visual Inspection Technology of Plate Quality Based on YOLO v5 Algorithm
LIU Fenghua1 ,XIE Guoxian2 ,XIAO Haonan2 ,YANG Liangsheng1 ,LI Jianfeng1
(1.Guangzhou KDT Machinery Co.,Ltd.,Guangzhou 510535,China; 2.Guangzhou Wangshi Software Technology Co.,Ltd.,Guangzhou 510535,China)
Abstract:For the automatic production of furniture panel,fast and accurate panel quality inspection system is indispensable. Based on YOLO v5 algorithm,the technology of computer-aided inspection of plate quality is studied. On the basis of machine learning algorithm,combined with the moving object capture algorithm as an auxiliary detection method,the panel recognition and location were realized. Then the panel size measurement is realized by image processing and contour recognition in OpenCV. The recognition rate of the algorithm is 98.33%. When recognizing the length and width dimensions of multi-color panels,the measured maximum dimension error is 2.23 mm,which can meet the needs of furniture panel production quality detection.
Keywords:machine vision;YOLO v5;edge detection;size detection
参考文献:
[1] KE Z N,ZHAO Q J,HUANG C H,et al. Detection of wood surface defects based on particle swarm-genetic hybrid algorithm [C]//2016 International Conference on Audio, Language and Image Processing (ICALIP).Shanghai:IEEE, 2017:375-379.
[2] 唐小松,庄宏金,罗平,等 . 一种板式家具板材 CCD 检 测装置:CN206146390U [P].2017-05-03.
[3] 龚建钊,华城,周恒,等 . 基于机器视觉的定制家居板 件在线尺寸检测系统研究 [J]. 中国人造板,2020,27(1):20- 22+27.
[4] 邓斌攸,潘云峰,池志强,等 . 基于机器视觉的家具板件 尺寸在线测量系统 [J]. 木材科学与技术,2021,35(2):63-67.
[5] GU J X,WANG Z H,KUEN J,et al. Recent Advances in Convolutional Neural Networks [J].Pattern Recognition,2018(77): 354-377.
[6] REDMON J,FARHADI A. YOLO9000:Better,Faster, Stronger [C]//2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).Honolulu:IEEE,2017:6517-6525.
[7] 高宗,李少波,陈济楠,等 . 基于 YOLO 网络的行人检测 方法 [J]. 计算机工程,2018,44(5):215-219+226.
[8] 黎洲,黄妙华 . 基于 YOLO_v2 模型的车辆实时检测 [J]. 中国机械工程,2018,29(15):1869-1874.
[9] LI Z X,ZHOU F Q. FSSD:Feature Fusion Single Shot Multibox Detector [J/OL].arXiv:1712.00960 [cs.CV].(2017-12-04). https://arxiv.org/abs/1712.00960v1.
[10] JOHANSON E,PALIBERG THAGMAN O. Fast Visual recognition of Scots pine boards using template matching [J].Computers and Electronics in Agriculture,2015(118):85-91.
作者简介:刘风华(1968—),男,汉族,北京人,副高级工 程师,硕士,研究方向:家具智能制造。