| 摘 要: 为避免铁路接触线异物影响火车的正常行驶,文章提出一种基于脉冲神经网络的模型对接触线异物进行检测。首先,基于正常和异常的接触线图像编码得到的深度特征之间存在差距,实现对接触线异物的有效检测;其次,通过倒残差结构搭建脉冲序列生成模块;最后,基于脉冲神经网络的编码器提取特征信息。实验结果表明,在接触线异物检测数据集上,该模型的准确率和F1分数分别为99.70%和99.70%。同时,在 CIFAR-10(Canadian Institute for Advanced Research-10)和CIFAR-100(Canadian Institute for Advanced Research-100)数据集上的对比实验中,模型的准确率分别达到91.16%和79.54%。综上所述,该模型具有较强的分类检测能力,能够更准确地检测出异常接触线图像。 |
| 关键词: 异物检测;铁路接触线;脉冲神经网络;倒残差结构;自注意力机制 |
|
中图分类号: TP391
文献标识码: A
|
| 基金项目: 河北省自然科学基金(F2022208002) |
|
| Research on Foreign Object Detection in Railway Contact Line Based on Spiking NeuralNetwork |
|
ZHANG Yongqiang1,3, LIU Jianzhang1, LI Xiangnan2,3
|
(1.School of Information Science and Engineering, Hebei University of Science and Technology, Shijiazhuang 050018, China;
2.Shijiazhuang Chunxiao Internet Information Technology Co., Ltd., Shijiazhuang 050061, China;
3.Hebei Technology Innovation Center of Intelligent IoT, Shijiazhuang 050018, China)
zyq@hebust.edu.cn; ljz_work1998@163.com; xiangnan.li@chunxiao.net
|
| Abstract: To prevent foreign objects on railway contact lines from affecting the normal operation of trains, this paper proposes a model based on the Spiking Neural Network for detecting foreign objects on contact lines. Firstly, the disparity between deep features encoded from normal and abnormal contact line images is utilized to achieve effective detection of foreign objects. Secondly, an inverted residual structure is employed to construct a spike sequence generation module. Finally, a feature extraction encoder based on the Spiking Neural Network is used to extract feature information. Experimental results demonstrate that on the contact line foreign object detection dataset, the proposed model achieves an accuracy and F1 score of 99. 70% and 99. 70% , respectively. Additionally, in comparative experiments on the CIFAR-10 (Canadian Institute for Advanced Research-10) and CIFAR-100 (Canadian Institute for Advanced Research-100) datasets, the accuracy of the model reaches 91.16% and 79.54% , respectively. In conclusion, the proposed model exhibits strong classification and detection capabilities, enabling more accurate identification of abnormal contact line images. |
| Keywords: foreign object detection; railway contact line; Spiking Neural Network; inverted residual structure; sel-f attention mechanism |
