基于迁移学习的高铁线缆扣件装配检测

The Assembly Detection of High-Speed Rail Cable Fastener Based on Transfer Learning

为了判断高铁线缆扣件的装配是否正确,这里采用一种基于迁移学习的卷积神经网络的算法对高铁线缆扣件装配进行检测。首先将预训练的网络与目标检测算法相结合,建立完整的装配检测网络,然后对制作好的数据集进行训练和测试。实验结果表明,相比传统对象识别的方法,该方法不仅提高了工件装配检测的准确度,还保证了工业检测中对实时性的要求。另外,由于卷积神经网络可以获取工件图像的深层特征,从而使得目标检测算法更加稳健,更能适应光照、灰尘等环境噪声的变化。

In order to judge whether the assembly of high-speed rail cable fasteners are correct,this paper uses a convolutional neural network based on transfer learning to detect the assembly of high-speed rail cable fasteners.First,the pre-trained network is combined with the target detection algorithm to build a complete assembly inspection network,and then the trained data set is trained and tested.The experimental results show that compared with the traditional method of object recognition,this method not only improves the accuracy of workpiece assembly inspection,but also ensures the real-time requirements in industrial inspection.In addition,convolutional neural network can obtain the deep features of the workpiece image,which makes the target detection algorithm more stable and more suitable for changes in environmental noise such as light and dust.