Research on Railway Track Foreign Object Intrusion Detection Based on Multi-Scale Feature Fusion

WANG Nan; HOU Tao; NIU Hongxia

doi:10.7652/xjtuxb202409014

您当前的位置：

首页 >

文章列表页 >

Research on Railway Track Foreign Object Intrusion Detection Based on Multi-Scale Feature Fusion

更新时间：2025-07-28

- Research on Railway Track Foreign Object Intrusion Detection Based on Multi-Scale Feature Fusion
- Vol. 58, Issue 9, Pages: 139-153(2024)
- 作者机构：
  
  兰州交通大学自动化与电气工程学院,兰州,730070
- 作者简介：
- 基金信息：
- DOI：10.7652/xjtuxb202409014
  CLC： U298;TP391.41
- Published：2024
- 稿件说明：
移动端阅览
WANG Nan, HOU Tao, NIU Hongxia. Research on Railway Track Foreign Object Intrusion Detection Based on Multi-Scale Feature Fusion[J]. 2024, 58(9): 139-153.
DOI：

WANG Nan, HOU Tao, NIU Hongxia. Research on Railway Track Foreign Object Intrusion Detection Based on Multi-Scale Feature Fusion[J]. 2024, 58(9): 139-153. DOI： 10.7652/xjtuxb202409014.

摘要

针对铁路轨道异物检测中不同尺度异物目标的检测易受复杂环境的影响

导致出现检测精度低及检测速度慢等问题

提出一种多尺度特征融合的铁轨异物检测(RMF-YOLO)算法。首先

设计并引入改进的卷积注意力模块(ICBAM)

结合YOLOv7特征提取网络

以增强复杂场景下的特征提取能力。其次

在所有高效层聚合网络模块中采用GhostConv替代普通卷积层

以降低计算复杂度

提高特征输出效率; 设计一种改进的加权双向特征金字塔网络N-BiFPN结构

加强多尺度特征融合能力

平衡不同层级特征信息

提高多尺度检测能力。最后

为进一步提升检测精度

采用WIoU损失函数结合动态非单调聚集机制

有效应对低质量锚框产生的梯度

提高模型对不同尺度异物检测的整体性能。实验结果表明:在自制的铁轨异物数据集上

RMF-YOLO算法减少了原网络模型的参数量

有效提升了模型的检测精度与检测速度

改善了漏检与误检问题

平均精度提升了5.5%

检测速度提升了5.88%

计算量减少了12.25%

能满足铁轨入侵异物检测中对检测精度和实时性的需求。

Abstract

To address the issues of low detection accuracy and slow detection speed caused by the influence of complex environments on detecting foreign objects of different scales on railroad tracks

this paper proposes a multi-scale feature fusion RMF-YOLO(railway multi-scale fusion YOLO)method for railroad track foreign object detection. Firstly

ICBAM

an improved convolutional attention module

is designed and introduced in combination with the YOLOv7 feature extraction network to enhance feature extraction capability in complex scenarios. Subsequently

GhostConv is adopted in all efficient layer aggregation network modules instead of regular convolutional layers to reduce computational complexity and enhance feature output efficiency.An improved weighted bi-directional feature pyramid network(N-BiFPN)structure is introduced to enhance multi-scale feature fusion

balance feature information across different levels

and improve multi-scale detection capabilities. Lastly

to further enhance detection accuracy

the WIoU loss function combined with the dynamic non-monotonic aggregation mechanism is used to effectively deal with the gradient generated by the low-quality anchor frames and to improve the overall performance of the model for the detection of foreign objects at different scales. The experimental results indicate that on the self-made railway track foreign object dataset

the RMF-YOLO algorithm reduces the parameter count of the original network model

effectively enhancing the model's detection accuracy and detection speed of the model and addressing issues related to missed and false detections. On average

the precision increased by 5.5%

the frames per second(FPS)improved by 5.88%

and the computational load decreased by 12.25%. These enhancements meet the requirements for both detection accuracy and real-time performance in railway track foreign object detection.

关键词

Keywords

references

RISTIC-DURRANT D, FRANKE M, MICHELS K. A review of vision-based on-board obstacle detection and distance estimation in railways [J]. Sensors, 2021, 21(10): 3452.

LEI Yang, HE Xiuxiu, YAO Jincao, et al. Breast tumor segmentation in 3D automatic breast ultrasound using mask scoring R-CNN [J]. Medical Physics, 2021, 48(1): 204-214.

LAW H, DENG Jia. CornerNet: detecting objects as paired keypoints [J]. International Journal of Computer Vision, 2020, 128(3): 642-656.

XU Renjie, LIN Haifeng, LU Kangjie, et al. A forest fire detection system based on ensemble learning [J]. Forests, 2021, 12(2): 217.

CAO Jinghao, ZHANG Junju, JIN Xin. A traffic-sign detection algorithm based on improved sparse R-CNN [J]. IEEE Access, 2021, 9: 122774-122788.

WANG Dashuai, LI Zhuolin, DU Xiaoqiang, et al. Farmland obstacle detection from the perspective of UAVs based on non-local deformable DETR [J]. Agriculture, 2022, 12(12): 1983.

GUO Baoqing, SHI Jiafeng, ZHU Liqiang, et al. High-speed railway clearance intrusion detection with improved SSD network [J]. Applied Sciences, 2019, 9(15): 2981.

汪世财, 谈东奎, 谢有浩, 等. 基于激光雷达点云密度特征的智能车障碍物检测与跟踪 [J]. 合肥工业大学学报(自然科学版), 2019, 42(10): 1311-1317.

WANG Shicai, TAN Dongkui, XIE Youhao, et al. Obstacle detection and tracking for intelligent vehicle based on density characteristics of point cloud using 3D lidar [J]. Journal of Hefei University of Technology(Natural Science), 2019, 42(10): 1311-1317.

徐岩, 陶慧青, 虎丽丽. 基于Faster R-CNN网络模型的铁路异物侵限检测算法研究 [J]. 铁道学报, 2020, 42(5): 91-98.

XU Yan, TAO Huiqing, HU Lili. Railway foreign body intrusion detection based on Faster R-CNN network model [J]. Journal of the China Railway Society, 2020, 42(5): 91-98.

张剑, 王等准, 莫光健, 等. 基于改进YOLOv3的高铁异物入侵检测算法 [J]. 计算机技术与发展, 2022, 32(2): 69-74.

ZHANG Jian, WANG Dengzhun, MO Guangjian, et al. High-speed rail foreign body intrusion detection algorithm based on improved YOLOv3 [J]. Computer Technology and Development, 2022, 32(2): 69-74.

HUSSAIN R, KARBHARI Y, IJAZ M F, et al. Revise-net: exploiting reverse attention mechanism for salient object detection [J]. Remote Sensing, 2021, 13(23): 4941.

胡行涛, 刘大明, 虞发桐. 基于FCN不确定性特征的铁路入侵异物检测算法 [J]. 计算机应用与软件, 2023, 40(4): 141-146.

HU Xingtao, LIU Daming, YU Fatong. Foreign intrusion object detection algorithm for railway intrusion based on FCN uncertainty [J]. Computer Applications and Software, 2023, 40(4): 141-146.

CAI Huiling, LI Fei, GAO Dianzhu, et al. Foreign objects intrusion detection using millimeter wave radar on railway crossings [C]//2020 IEEE International Conference on Systems, Man, and Cybernetics(SMC).Piscataway, NJ, USA: IEEE, 2020: 2776-2781.

HE Deqiang, QIU Yefeng, MIAO Jian, et al. Improved mask R-CNN for obstacle detection of rail transit [J]. Measurement, 2022, 190: 110728.

WANG Yao, YU Zujun, ZHU Liqiang. Intrusion detection for high-speed railways based on unsupervised anomaly detection models [J]. Applied Intelligence, 2023, 53(7): 8453-8466.

管岭, 贾利民, 谢征宇. 融合注意力机制的轨道入侵异物检测轻量级模型研究 [J]. 铁道学报, 2023, 45(5): 72-81.

GUAN Ling, JIA Limin, XIE Zhengyu. Research on lightweight model for railway intrusion detection integrating attention mechanism [J]. Journal of the China Railway Society, 2023, 45(5): 72-81.

HE Deqiang, YANG Yanjie, CHEN Yanjun, et al. An integrated optimization model of metro energy consumption based on regenerative energy and passenger transfer [J]. Applied Energy, 2020, 264: 114770.

WANG C Y, BOCHKOVSKIY A, LIAO H-Y M. YOLOv7: trainable bag-of-freebies sets new state-of-the-art for real-time object detectors [C]//2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR).Piscataway, NJ, USA: IEEE, 2023: 7464-7475.

HUANG Peile, WANG Shenghuai, CHEN Jianyu, et al. Lightweight model for pavement defect detection based on improved YOLOv7 [J]. Sensors, 2023, 23(16): 7112.

WOO S, PARK J, LEE J Y, et al. CBAM: convolutional block attention module [C]//Computer Vision-ECCV 2018. Cham, Germany: Springer International Publishing, 2018: 3-19.

HAN Kai, WANG Yunhe, TIAN Qi, et al. GhostNet: more features from cheap operations [C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR).Piscataway, NJ, USA: IEEE, 2020: 1577-1586.

ZHAO Xiaotong, LI Wei, ZHANG Yifan, et al. Aggregated residual dilation-based feature pyramid network for object detection [J]. IEEE Access, 2019, 7: 134014-134027.

YU Haifeng, LI Xinbin, FENG Yankai, et al. Multiple attentional path aggregation network for marine object detection [J]. Applied Intelligence, 2023, 53(2): 2434-2451.

SONG Zhiwei, HUANG Xinbo, JI Chao, et al. Intelligent identification method of hydrophobic grade of composite insulator based on efficient GA-YOLO former network [J]. IEEJ Transactions on Electrical and Electronic Engineering, 2023, 18(7): 1160-1175.

GAO Jinfeng, CHEN Yu, WEI Yongming, et al. Detection of specific building in remote sensing images using a novel YOLO-S-CIOU model. case: gas station identification [J]. Sensors, 2021, 21(4): 1375.

ZHAO Qing, WEI Honglei, ZHAI Xianyi. Improving tire specification character recognition in the YOLOv5 network [J]. Applied Sciences, 2023, 13(12): 7310.

张开放, 苏华友, 窦勇. 一种基于混淆矩阵的多分类任务准确率评估新方法 [J]. 计算机工程与科学, 2021, 43(11): 1910-1919.

ZHANG Kaifang, SU Huayou, DOU Yong. A new multi-classification task accuracy evaluation method based on confusion matrix [J]. Computer Engineering Science, 2021, 43(11): 1910-1919.

GE Zheng, LIU Songtao, WANG Feng, et al. YOLOX: exceeding YOLO series in 2021 [EB/OL].(2021-08-06)[2023-11-12]. https://arxiv.org/abs/2107.08430.

SU Xiaodong, HU Jianxing, CHEN Linzhou, et al. Research on real-time dense small target detection algorithm of UAV based on YOLOv3-SPP [J]. Journal of the Brazilian Society of Mechanical Sciences and Engineering, 2023, 45(9): 488.

TIAN Zhi, SHEN Chunhua, CHEN Hao, et al. FCOS: a simple and strong anchor-free object detector [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022, 44(4): 1922-1933.

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

Image Super Resolution Algorithm Based on Receptive Field Optimization and Progressive Feature Fusion

Nonlinear Multi-Scale Contrast Enhancement for X-Ray Chest Radiography

A Corner Detection Algorithm Based on Multi-Scale Directional Differential Ratio

A Multi-Class Cost Sensitivity AdaBoost Algorithm Using Multi-Class Cost Exponential Loss Function

A Method of Image Content Matching with Lattice Closeness and Multi-Feature Fusion

Related Author

DA Feipeng

GAI Shaoyan

WU Hongwu

牟轩沁

张敏

水鹏朗

王富平

贾琪

Related Institution

School of Automation, Southeast University

Key Laboratory of Measurement and Control of Complex Systems of Engineering, Ministry of Education, Southeast University

Institute of Image Processing and Pattern Recognition, Xi'an Jiaotong University

National Key Lab of Radar Signal Processing, Xidian University

Institute of Air Defense and Anti-Missile, Air Force Engineering University

⁰