摘要
传统的陶瓷衬垫码片识别主要依赖于人工,由人工手动摆正方向错误的码片,识别效率低,还有可能形成错检、漏检,造成后续工程问题。对此,提出一种基于深度学习的陶瓷衬垫码片识别方法。这一方法通过预训练的Mask区域卷积神经网络对已标记的样本进行试验,使用Res Net-50网络进行特征提取,调节神经网络的结构,并计算求解。在研究中,通过将两种检测类别标记相结合的方式,解决了错位码片的识别问题。试验表明,基于深度学习的陶瓷衬垫码片识别方法能够有效识别摆放错位的码片,准确度高。
The traditional identification of the code chip on the ceramic liner mainly relies on manual work.Manual aligning of the wrongly oriented chips has low identification efficiency,and may lead to error detection and missed detection,which may cause subsequent engineering problems.In this regard,an identification solution for the code chip on the ceramic liner based on deep learning was proposed.This method uses the pre-trained Mask R-CNN to test the labeled samples,uses the Res Net-50 network for feature extraction,adjusts the structure of the neural network,and calculates the solution.In the research,the problem of identifying misplaced chips was solved by combining the two detected category marks.Tests show that the identification solution for the code chip on the ceramic liner based on deep learning can effectively identify misplaced chips with high accuracy.
出处
《机械制造》
2021年第5期45-49,共5页
Machinery
关键词
深度学习
陶瓷衬垫
码片
识别
研究
Deep Learning
Cement Liner
Code Chip
Recognition
Research
作者简介
魏智锋(1994-),男,硕士研究生,主要研究方向为机器视觉;肖书浩(1962-),男,副教授,主要研究方向为机器视觉。
