摘要
针对大型多面复杂结构件表面打磨、喷涂等处理对自动化、智能化的需求,传统目标检测具有局限性、小目标检测准确度低、速率慢等缺点,本文结合计算机视觉提出一种改进YOLOv3算法的工件喷涂质量检测方法,结合实际工控环境需求,构建喷涂数据集,实现对工件表面迷彩喷涂质量的检测.首先,使用K-means++对Anchor重新聚类生成适合本文的锚框尺寸;通过图像增强技术对数据集进行增强,并对YOLOv3的Darknet-53网络结构进行改进,在保证计算准确度的同时提高效率.实验结果表明,本文提出的改进YOLOv3算法,能够准确快速的对迷彩喷涂缺陷进行定位.
For large surface more complicated structure surface polishing,spraying processing demand for automatic and intelligent,traditional target detection limits,small target detection rate of low accuracy and slow shortcomings,this paper puts forward an improved YOLOv3 computer vision algorithm of workpiece spraying quality detection method,combined with the actual engineering environment demand,building coating data sets,for detecting camouflage coating on the surface of a workpiece quality.First,k-means++is used to re-cluster Anchor to generate the anchor box size appropriate for this article.The data set was enhanced through image enhancement technology,and the Darknet-53 network structure of YOLOv3 was improved to ensure the accuracy of calculation while improving the efficiency.Experimental results show that the improved YOLOv3 algorithm proposed in this paper can accurately and quickly locate the defects in camouflage spraying.
作者
顾旭
郭锐锋
王鸿亮
张晓星
GU Xu;GUO Rui-feng;WANG Hong-liang;ZHANG Xiao-xing(University of Chinese Academy of Sciences,Beijing 100049,China;Shenyang Institute of Computing Technology,Chinese Academy of Sciences,Shenyang 110168,China)
出处
《小型微型计算机系统》
CSCD
北大核心
2022年第2期343-348,共6页
Journal of Chinese Computer Systems
基金
辽宁省工业重大专项项目(2019JH1/10100007)资助
沈阳市高层次人才创新创业团队项目(2019-SYRCCX-C-08)资助。