摘 要：针对列车故障检测效率低的问题，提出一种基于 MobielNet 的移动端列车图像故障检测算法。首先，在 MobileNet中引入注意力卷积块和 Ghost 模块，用以提升网络的学习能力。其次，使用残差聚合网络获取多层次的特征图。最后，将该模型移植到移动端设备上完成列车故障检测任务。实验结果表明，该算法的平均精度均值达到了 85.35%，与 YOLOv3-Tiny、YOLOv4-Tiny、YOLOX、YOLOv5 相比，mAP 分别提高了 8.83%、5.49%、7.89%、5.31%，并且 FED 拥有更低检测延迟。
关键词：列车故障检测；目标识别；MobileNet；移动设备；注意力机制
DOI:10.19850/j.cnki.2096-4706.2023.06.012
中图分类号：TP391.4 文献标识码：A 文章编号：2096-4706（2023）06-0046-05
Mobile Terminal Train Image Fault Detection Algorithm Based on MobielNet
ZHOU Peng, ZHANG Longxin
(Hunan University of Technology, Zhuzhou 412007, China)
Abstract: To solve the problem of low efficiency of train fault detection, mobile terminal train image fault detection algorithm based on MobielNet is proposed. First, attention convolution block and Ghost module are introduced into MobileNet to improve the learning ability of the network. Secondly, residual aggregation network is used to obtain multi-level feature map. Finally, the model is transplanted to the mobile terminal equipment to complete the train fault detection task. The experimental results show that the average accuracy of the algorithm reaches 85.35%. Compared with YOLOv3-Tiny, YOLOv4-Tiny, YOLOX and YOLOv5, mAP improves 8.83%, 5.49%, 7.89% and 5.31% respectively, and FED has lower detection delay.
Keywords: train fault detection; target recognition; MobileNet; mobile device; attention mechanism
参考文献：
[1] EL-SAADAWY H，TANTAWI M，Shedeed H A，et al. A Two-Stage Method for Bone X-Rays Abnormality Detection Using MobileNet Network [C]//AICV: The International Conference on Artificial Intelligence and Computer Vision.Cairo：Springer，2020，1153：372-380.
[2] MAO Q C，SUN H M，LIU Y B，et al. Mini-YOLOv3:Real-Time Object Detector for Embedded Applicatio [J].IEEE Access， 2019，7：133529-133538.
[3] BOCHKOVSKIY A，WANG C Y，LIAO H Y M. YOLOv4: Optimal Speed and Accuracy of Object Detection [J/OL]. arXiv:2004.10934 [cs.CV].[2022-09-28].https://arxiv.org/abs/2004.10934.[4] ZHU X K，LYU S C，WANG X，et al. TPH-YOLOv5:
Improved YOLOv5 Based on Transformer Prediction Head for Object Detection on Drone-captured Scenarios [C]//2021 IEEE/CVF International Conference on Computer Vision Workshops（ICCVW）. Montreal：IEEE，2021：2778-2788.
[5] PANBOONYUEN T，THONGBAI S，WONGWEERANIMIT W，et al. Object Detection of Road Assets Using Transformer-Based YOLOX with Feature Pyramid Decoder on Thai Highway Panorama [J/ OL].Information，2022，13（1）[2022-09-28].https://doi.org/10.3390/info13010005.
[6] LI Y，LIU Y，CUI W G，et al. Epileptic Seizure Detection in EEG Signals Using a Unified Temporal-Spectral Squeeze-and-Excitation Network [J].IEEE Transactions on Neural Systems and Rehabilitation Engineering，2020，28（4）：782-794.
作者简介：周鹏（1997—），男，汉族，湖南常德人，硕士在读，研究方向：基于深度学习的列车识别方法；张龙信（1983—），男，汉族，湖南株洲人，副教授，博士，研究方向：高性能计算、机器学习。