基于极限学习机的玻璃瓶口缺陷检测方法研究
摘要：针对目前玻璃空瓶回收再生产过程中造成瓶口缺陷破损的在线实时检测难题，提出一种基于极限学习机（Extreme Learning Machine， ELM）的检测算法。首先对采集的瓶口进行预处理，通过研究表面缺陷，提取灰度方差等6种表面特征。然后运用遗传算法对极限学习机的输入层层的阈值和权值进行优化，提高算法的检测精度。最后现场选取569个样本对所设计ELM分类器进行训练学习与测试，并与LVQ算法、BP分类器对比实验。结果表明该算法能够满足对机器视觉检测系统缺陷检测高速高精度的要求。
关键词：缺陷检测；机器视觉；特征提取；极限学习机
中图分类号：TP391.4文献标识码：A
Abstract：A novel defect detection method based on Extreme Learning Machine was proposed for beer bottle mouth， which tackles with the problem of beer online realtime defect detection in recycling and reproduction process. The proposed method consists of the following steps. First， the bottle mouth is preprocessed by researching on the characteristics of surface defect bottle mouth， which extracts six kinds of surface features such as gray scale variance. Then， to improve the detection accuracy， we optimize Extreme Learning Machine （ELM） input and output layers of threshold and weight by using genetic algorithm. Finally， 569 samples from experimental test platform are selected to design the ELM classifier， and experimental results are compared with LVQ algorithm and BP algorithm. Experimental results show that the proposed ELM based classifier is able to obtain much higher speed and higher detection accuracy， which can meet the requirements of the production enterprise for machine vision system.
Key words：defect detection； machine vision； feature extraction； extreme learning machine
1引言
质检行业面临劳动力供给不足和劳动力成本上升的压力，采用自动化视觉机器代替人工操作岗位，是解决制造业用工问题的最有效途径。机器视觉检测已在集成...
|
== 试读已结束，如需继续阅读敬请充值会员 ==
|
|
本站文章均为原创投稿，仅供下载参考，付费用户可查看完整且有格式内容！
(费用标准：38元/2月，98元/2年，微信支付秒开通！) |
| 升级为会员即可查阅全文 。如需要查阅全文，请 免费注册 或 登录会员 |
|
