范涛,朱青,王耀南,周显恩,刘远强.空瓶检测机器人瓶底缺陷检测方法研究[J].电子测量与仪器学报,2017,31(9):1394-1401
空瓶检测机器人瓶底缺陷检测方法研究
Research on detection method of bottle bottom defects based on empty bottle detection robot system
  
DOI:10.13382/j.jemi.2017.09.007
中文关键词:  机器视觉  瓶底定位  缺陷检测  支持向量机
英文关键词:machine vision  bottle bottom location  defect detection  support vector machine
基金项目:国家自然科学基金(61573134)资助项目
作者单位
范涛 湖南大学 电气与信息工程学院 机器人视觉感知与控制技术国家工程实验室长沙410082 
朱青 湖南大学 电气与信息工程学院 机器人视觉感知与控制技术国家工程实验室长沙410082 
王耀南 湖南大学 电气与信息工程学院 机器人视觉感知与控制技术国家工程实验室长沙410082 
周显恩 湖南大学 电气与信息工程学院 机器人视觉感知与控制技术国家工程实验室长沙410082 
刘远强 佛山市湘德智能科技有限公司佛山528000 
AuthorInstitution
Fan Tao National Engineering Laboratory for Robot Visual Perception and Control Technology, College of Electrical and Information Engineering, Hunan University, Changsha 410082, China 
Zhu Qing National Engineering Laboratory for Robot Visual Perception and Control Technology, College of Electrical and Information Engineering, Hunan University, Changsha 410082, China 
Wang Yaonan National Engineering Laboratory for Robot Visual Perception and Control Technology, College of Electrical and Information Engineering, Hunan University, Changsha 410082, China 
Zhou Xianen National Engineering Laboratory for Robot Visual Perception and Control Technology, College of Electrical and Information Engineering, Hunan University, Changsha 410082, China 
Liu Yuanqiang Foshan Xiangde Intelligent Technology Co. Ltd., Foshan 52800, China 
摘要点击次数: 350
全文下载次数: 243
中文摘要:
      针对当前瓶底圆心定位方法精度不高、瓶底防滑纹区域缺陷易误检等问题,利用瓶底防滑纹的几何特征,提出一种改进的基于变权重随机圆拟合的瓶底定位算法,首先采用重心法对瓶底圆心进行快速预定位,再采用变权重随机圆拟合法实现瓶底精定位。然后检测瓶底图像疑似缺陷区域,并提取区域面积、轮廓长度、圆形度、灰度方差和灰度均值等特征,采用支持向量机算法进行分类决策,检测出缺陷。实验表明,瓶底定位误差小于6个像素,缺陷检测准确率为92.7%,基本满足实际生产精度的要求。
英文摘要:
      Aiming at the problem that the method of positioning the bottle bottom center is not accurate and the results for detecting the anti skid grain areas of bottom are unreliable, by taking advantage of the geometric features of anti skid grain areas on the bottom of the bottle, a localization algorithm based on variable weight random circle fitting the bottom is proposed in the paper. First, the bottom center is quickly pre positioned by gravity method, then the random variable weight circle fitting method is used to realize the precise positioning. Finally, the suspected defect region of the bottle bottom image is detected, and area, contour length, average gray, gray variance and circularity are extracted, then the support vector machine is used for classification and the defect is detected. The experiment results show that the positioning error of this method is less than 6 pixels, and the detect accuracy is 92.7%. It basically meets the actual production requirements.
查看全文  查看/发表评论  下载PDF阅读器