摘要
针对传统瓶盖字符人工检测误判率高、检测效率低的问题,研发一套基于机器视觉的瓶盖曲面字符缺陷检测系统。首先,对检测系统结构与视觉信息采集原理进行介绍;在此基础上,通过对所采集视觉信息进行滤波降噪处理,减少视觉信息背景噪声。然后对字符分割使用垂直投影和连通域分析法,再用特征点提取法和模板匹配对曲面字符进行识别。最后与射频识别(RFID)读码器所读取数字码比对,判定并剔除缺陷瓶盖,以此取代目前的人工检测方法。实验表明,该系统能够有效地对瓶盖字符进行缺陷检测,识别字符准确率为99.97%,处理一副图像时间为200ms,能够满足工业现场实际要求。
Aiming at the problem of high erroneous judgment rate and low detection efficiency of the traditional cap character artificial detection, a curved surface character defect detection system based on machine vision has been developed. Firstly, the structure of the detection system and the principle of visual information acquisition are introduced, on the basis of which, the collected visual information is processed by filtering and noise reduction, so the visual information background noise could been well reduced. For character segmentation, the method of vertical projection and connected domain analysis were adopted, after that, the surface character can be recognized by feature points extraction method and template matching, finally, compared with the digital code read by the RFID reader,the defect cap would be distinguished and eliminated, so the current manual detection method can be replaced. Experiment results show that the system can quickly and effectively detect the defect of cap character, recognition accuracy rate is 99.97% , and process the image time is 200ms, which can meet the practical requirements of industrial field.
作者
郭康
黄丹平
王磊
汪俊其
徐同旭
Guo Kang Huang Danping Wang Lei Wang Junqi Xu Tongxu(College of Mechanical Engineering, Sichuan University of Science ＆ Engineering, Zigong 643000, Sichuan, China Artificial Intelligence Key Laboratory of Sichuan Province, Zigong 643000, Sichuan, China Siehuan Provincial Key Lab of Process Equipment and Control, Zigong 643000, Sichuan, China)
出处
《现代制造工程》
CSCD
北大核心
2016年第12期122-127,共6页
Modern Manufacturing Engineering
基金
过程装备与控制工程四川省高校重点实验室项目(GK201209)
关键词
机器视觉
曲面字符
缺陷检测
垂直投影
模板匹配
machine vision
curved surface character
defect detection
vertical projection
template matching
作者简介
郭康，硕士研究生，主要研究方向：机器视觉．
黄丹平．通讯作者，副教授，主要研究方向：测控技术。E-mail：984554142@qq.com
