摘要
针对传统的空心胶囊缺陷检测多用于人工抽样检测,无论是时间成本还是人力成本都较高,并且带有一定的主观性的问题,研究分析了胶囊缺陷的特点,设计并搭建了胶囊制造缺陷在线检测实验平台,设计并开源了药用空心胶囊凹陷缺陷数据集,提出了一种基于Mask R-CNN算法的胶囊缺陷检测识别方法,使用ResNet作为特征提取网络,ROIAlign对特征图ROI提取是完成像素级的对齐,能很好地解决胶囊缺陷的检测难题。经过实验验证,基于Mask R-CNN的胶囊缺陷检测识别的正确率较高。
Traditional hollow capsule defect detection often adopts the manual sampling detection method,both the time cost and the labor cost are high,and there are certain subjective problems.The characteristics of capsule defects are studied and analyzed.The experimental platform for online detection of capsule manufacturing defects is designed and built.The data set of medicinal hollow capsule depression defect is designed and opened source.A capsule defect detection and recognition method based on Mask R-CNN algorithm is proposed.The method uses ResNet as the feature extraction network.ROIAlign extracts the feature map ROI to complete pixel-level alignment,which can solve the problem of capsule defect detection.After the experimental verification,the correct rate of capsule defect detection based on Mask R-CNN is high.
作者
段仲静
李少波
胡建军
杨静
王铮
DUAN Zhongjing;LI Shaobo;HU Jianjun;YANG Jing;WANG Zheng(Key Laboratory of Advanced Manufacturing Technology of Ministry of Education,Guizhou University,Guiyang 550025,China;School of Mechanical Engineering,Guizhou University,Guiyang 550025,China)
出处
《无线电工程》
2020年第10期857-862,共6页
Radio Engineering
基金
国家自然科学基金资助项目(51475097,91746116)。
作者简介
段仲静,男,(1992—),就读于贵州大学机械工程专业,硕士研究生,主要研究方向:产品在线无损质量检测;通讯作者:李少波,男,(1973—),博士,教授,主要研究方向:智能制造、大数据等;胡建军,男,(1973—),博士,教授,主要研究方向:大数据、深度学习;杨静,男,(1992—),博士研究生,主要研究方向:机器视觉、智能制造、机器人;王铮,男,(1994—),硕士研究生,主要研究方向:机械产品在线无损质量检测。
