摘要
基于机器视觉自动检测方式,提出了一种多任务卷积神经网络检测螺栓异常的方法。该方法相比传统图像比对检测故障方式,准确率及检测效率均有较大提升。
On the basis of the visual automatic inspection method by machine,a method for inspection of abnormal screw bolts for rail vehicles based upon multitask convolutional neural network is put forward.The method has improved greatly in accuracy and inspection efficiency comparing with the trouble inspection method with traditional image comparison.
作者
王勇
袁啸阳
陈铎
厉承臻
李健
崔伟
李振宝
WANG Yong;YUAN Xiaoyang;CHEN Duo;LI Chengzhen;LI Jian;CUI Wei;LI Zhenbao(CRRC Qingdao Sifang Rolling Stock Research Institute Co.,Ltd.,Qingdao 266031,China;不详)
出处
《铁道车辆》
2020年第5期29-32,I0003,共5页
Rolling Stock
关键词
多任务卷积神经网络
螺栓异常
图像对比
multitask convolutional neural network
screw abnormality
image comparison
