复杂天气下铁轨异物入侵目标检测研究

魏雨桐; 侯涛; 牛宏侠

doi:null

您当前的位置：

首页 >

文章列表页 >

复杂天气下铁轨异物入侵目标检测研究

更新时间：2026-04-30

- 复杂天气下铁轨异物入侵目标检测研究
- Research on Target Detection of Foreign Objects Intruding onto Railway Tracks under Complex Weather Conditions
- 西安交通大学学报 2026年
- 作者机构：
  
  兰州交通大学自动化与电气工程学院，730070，兰州
- 作者简介：
- 基金信息：
- DOI：
  中图分类号： TP391.41;U298
- 收稿：2026-03-13，
  
  修回：2026-04-30，
  
  录用：2026-04-30，
- 稿件说明：
移动端阅览
魏雨桐, 侯涛, 牛宏侠. 复杂天气下铁轨异物入侵目标检测研究[J/OL]. 西安交通大学学报, 2026.

WEI Yutong, HOU Tao, NIU Hongxia. Research on Target Detection of Foreign Objects Intruding onto Railway Tracks under Complex Weather Conditions[J/OL]. JOURNAL OF XI’AN JIAOTONG UNIVERSITY, 2026.
魏雨桐, 侯涛, 牛宏侠. 复杂天气下铁轨异物入侵目标检测研究[J/OL]. 西安交通大学学报, 2026. DOI：

WEI Yutong, HOU Tao, NIU Hongxia. Research on Target Detection of Foreign Objects Intruding onto Railway Tracks under Complex Weather Conditions[J/OL]. JOURNAL OF XI’AN JIAOTONG UNIVERSITY, 2026. DOI：

摘要

为保障复杂天气下列车运行安全，针对复杂天气下，铁轨异物入侵目标检测清晰度低，容易出现误检漏检及参数量大与推理速度慢等问题，提出基于改进YOLOv11n的复杂天气铁路轨道异物入侵检测算法RCW-YOLO。首先，新建雨雾感知动态融合门控，设计雨雾浓度估计器模块，提高检测精度。然后在检测头DynamicHead中新增雨雾注意力分支，并在卷积偏移量中融入雨雾浓度系数，减少误检率和漏检率。其次，在骨干和颈部网络中使用注意力增强多子带融合小波池化模块替代YOLOv11n现有卷积模块，在保持检测效果的同时减少参数量，提高推理速度。最后，引入新设计的损失函数EW-MPDIoU替代原损失函数CIOU，提高检测精度。实验结果表明，RCW-YOLO的处理结果的平均精度为84.6%，较YOLOv11n提升了9.2%，参数量下降了18.15%。与主流目标检测算法相比，RCW-YOLO的处理结果的平均精度较YOLOv9t、YOLOv10n分别提升了11.3%、11.3%，参数量较YOLOv9t、YOLOv10n分别下降了21.63%、25.34%，相比最新的YOLOv12、YOLOv13，平均精度分别提高了7.8%、7.0%，参数量分别下降了22.18%、19.93%。

Abstract

To ensure the safety of train operation under complex weather conditions

considering the low clarity of target detection for foreign objects invading the railway tracks under such conditions

and the problems of frequent false detections

missed detections

large parameter quantity and slow inference speed

a complex weather railway track foreign object intrusion detection algorithm RCW-YOLO(Railway Complex Weather-YOLO) based on improved YOLOv11n is proposed. Firstly

a rain and fog perception dynamic fusion gating is established

and a rain and fog concentration estimator module is designed to improve the detection accuracy. Then

a rain and fog attention branch is added to the detection head DynamicHead

and the rain and fog concentration coefficient is integrated into the convolution offset to reduce the false detection rate and missed detection rate. Secondly

the attention-enhanced multi-subband fusion wavelet pooling module is used instead of the existing convolution module in the backbone and neck networks to replace the YOLOv11n

maintaining the detection effect while reducing the parameter quantity and improving the inference speed. Finally

the newly designed loss function Edge-Weighted Minimum Point Distance Intersection over Union (EW-MPDIoU) is introduced to replace the original loss function CIOU

improving the detection accuracy. Experimental results show that the mAP@0.5 of the processing results of RCW-YOLO is 84.6%

which is 9.2% higher than that of YOLOv11n

and the parameter quantity is 18.15% lower than that of YOLOv11n. Compared with mainstream object detection algorithms

the mAP@0.5 of the processing results of RCW-YOLO is 11.3% higher than that of YOLOv9t and 11.3% higher than that of YOLOv10n

and the parameter quantity is 21.63% lower than that of YOLOv9t and 25.34% lower than that of YOLOv10n. Compared with the latest YOLOv12 and YOLOv13

the mAP@0.5 is 7.8% higher and the parameter quantity is 22.18% lower.

关键词

Keywords

references

徐展 . 基于轻量化目标检测的铁路周界异物入侵检测方法研究 [D ] . 沈阳工业大学 , 2025 .

XU Z . Research on Railway Perimeter Foreign Object Intrusion Detection Method Based on Lightweight Object Detection [D ] . Shenyang University of Technology , 2025 .

赵永超 , 丁东 , 王汉策 , 等 . 基于无人值守视觉传感系统的铁路边坡侵限落石智能监测方法 [J ] . 科技与创新 , 2025 ,( 20 ): 77 - 79+83 .

ZHAO Y , DING D , WANG H , et al . Intelligent Monitoring Method for Railway Slope Encroaching Rockfall Based on Unattended Visual Sensing System [J ] . Science and Innovation , 2025 ,( 20 ): 77 - 79+83 .

HOU P , CHEN L . Rain-fog detection via spatial adaptive deformable network with multi-scale feature preservation and task-aware dynamic calibration [J ] . Engineering Research Express , 2025 , 7 ( 3 ): 035248 - 035248 .

朱思远 . 基于自适应变化检测和自监督机制的异物侵限检测方法研究 [D ] . 西安交通大学 , 2022 .

ZHU S . Research on Foreign Object Intrusion Detection Method Based on Adaptive Change Detection and Self-supervision Mechanism [D ] . Xi'an Jiaotong University , 2022 .

杨伟 , 王帅 , 吴佳奇 , 等 . 从局部到全局的零参考低照度图像增强方法 [J ] . 西安交通大学学报 , 2024 , 58 ( 04 ): 158 - 169 .

YANG W , WANG S , WU J , et al . A Zero-Reference Low-Illumination Image Enhancement Method from Local to Global [J ] . Journal of Xi'an Jiaotong University , 2024 , 58 ( 04 ): 158 - 169 .

高修强 , 余星阳 , 刘伯鹍 , 等 . 基于改进Mask R-CNN的轨道交通异物入侵模型研究 [J ] . 自动化与仪表 , 2025 , 40 ( 06 ): 111 - 115 .

GAO X , YU X , LIU B , et al . Research on the Rail Transit Foreign Object Intrusion Model Based on Improved Mask R-CNN [J ] . Automation and Instrumentation , 2025 , 40 ( 06 ): 111 - 115 .

陈俊明 . 面向铁路异物入侵检测的图像数据扩增方法研究 [D ] . 北京 : 北京交通大学 , 2023 .

CHEN J . Research on Image Data Augmentation Methods for Railway Foreign Object Intrusion Detection [D ] . Beijing : Beijing Jiaotong University , 2023 .

柳长源 , 夏苏辉 , 兰朝凤 . 恶劣天气下的航拍车辆目标检测算法 [J/OL ] . 北京航空航天大学学报 , 1 - 14 [ 2026-01-14 ] .

LIU C , XIA S , LAN C . Algorithm for Detecting Aerial Vehicle Targets under Severe Weather Conditions [J/OL ] . Journal of Beijing University of Aeronautics and Astronautics , 1 - 14 [ 2026-01-14 ] .

王诚 , 谢立云 , 李坤 . 基于改进YOLOv11的无人机雾天车辆目标检测算法研究 [J/OL ] . 南京邮电大学学报(自然科学版) , 1 - 9 [ 2026-03-11 ] .

WANG C , XIE L , LI K . Research on UAV Vehicle Target Detection Algorithm in Foggy Weather Based on Improved YOLOv11 [J/OL ] . Journal of Nanjing University of Posts and Telecommunications (Natural Science Edition) , 1 - 9 [ 2026-03-11 ] .

王井阳 , 徐勇超 , 张波 , 等 . 改进YOLOv10n的轻量化道路裂缝检测模型 [J/OL ] . 郑州大学学报(工学版) , 1 - 8 [ 2026-03-11 ] .

WANG J , XU Y , ZHANG B , et al . Lightweight road crack detection model based on improved YOLOv 10 n[J/OL ] . Journal of Zhengzhou University (Engineering Science) , 1 - 8 [ 2026-03-11 ] .

郭家虎 , 何磊 . DEL-YOLO:低照度轻量级煤矿输送带异物检测 [J/OL ] . 电子测量与仪器学报 , 1 - 12 [ 2026-01-14 ] .

GUO J , HE L . DEL-YOLO: Lightweight Low-Illumination Coal Conveyor Belt Foreign Object Detection [J/OL ] . Journal of Electronic Measurement and Instrumentation , 1 - 12 [ 2026-01-14 ] .

孙修宇 , 毕楠 . 基于改进YOLO11的低光照环境目标检测算法 [J ] . 激光与光电子学进展 , 2025 , 62 ( 22 ): 426 - 436 .

SUN X , BI N . Target Detection Algorithm in Low-Illumination Environments Based on Improved YOLO11 [J ] . Laser and Optoelectronic Progress , 2025 , 62 ( 22 ): 426 - 436 .

RUAN C , HONG L , GAO J . Marine-YOLO: A high-precision object detection algorithm for complex maritime environments [J ] . Applied Ocean Research , 2026 , 169104994 - 104994 .

张瑞芳 , 段世纪 , 刘占占 , 等 . MSDE-YOLO：低照度图像多尺度特征聚合目标检测算法 [J/OL ] . 电光与控制 , 1 - 9 [ 2026-01-14 ] .

ZHANG R , DUAN S , LIU Z , et al . MSDE-YOLO: Multi-scale Feature Aggregation for Target Detection in Low-Illumination Images [J/OL ] . Opto-Electronics and Control , 1 - 9 [ 2026-01-14 ] .

周强 . 低光图像目标检测的优化研究 [J/OL ] . 智能计算机与应用 , 1 - 17 [ 2026-01-14 ] .

ZHOU Q . Optimization Research on Target Detection in Low-Light Images [J/OL ] . Intelligent Computer and Applications , 1 - 17 [ 2026-01-14 ] .

王轩楷 , 潘广贞 , 陈新 . 改进YOLO11n的水下目标检测 [J ] . 计算机系统应用 , 2025 , 34 ( 10 ): 76 - 85 .

WANG X , PAN G , CHEN X . Improved YOLO11n for Underwater Object Detection [J ] . Computer Systems Applications , 2025 , 34 ( 10 ): 76 - 85 .

牛为华 , 郭迅 . 融合分块注意力与小波特征聚合的遥感图像目标检测算法 [J/OL ] . 智能系统学报 , 1 - 12 [ 2026-03-11 ] .

NIU W , GUO X . Remote Sensing Image Target Detection Algorithm Combining Block Attention and Wavelet Feature Aggregation [J/OL ] . Journal of Intelligent Systems , 1 - 12 [ 2026-03-11 ] .

王楠 , 侯涛 , 牛宏侠 . 多尺度特征融合的铁轨异物入侵检测研究 [J ] . 西安交通大学学报 , 2024 , 58 ( 09 ): 139 - 153 .

WANG N , HOU T , NIU H . Research on Railway Track Foreign Object Intrusion Detection with Multi-scale Feature Fusion [J ] . Journal of Xi'an Jiao tong University , 2024 , 58 ( 09 ): 139 - 153 .

张凡 , 张鹏超 , 王磊 , 等 . 基于YOLOv5s的轻量化朱鹮检测算法研究 [J ] . 西安交通大学学报 , 2023 , 57 ( 01 ): 110 - 121 .

ZHANG F , ZHANG P , WANG L , et al . Research on Lightweight Red-crowned Crane Detection Algorithm Based on YOLOv5s [J ] . Journal of Xi'an Jiaotong University , 2023 , 57 ( 01 ): 110 - 121 .

侯涛 , 席超 , 牛宏侠 . 复杂环境下铁路轨道异物入侵轻量化检测研究 [J/OL ] . 北京交通大学学报 , 1 - 14 [ 2026-04-15 ] .

HOU T , XI C , NIU H . Research on Lightweight Detection of Foreign Objects Intruding into Railway Tracks in Complex Environments [J/OL ] . Journal of Beijing Jiaotong University , 1 - 14 [ 2026-04-15 ] .

郝芝建 , 牟舵 , 王留毅 , 等 . 基于改进YOLOv10的隧道渗漏水智能检测方法 [J/OL ] . 华南理工大学学报(自然科学版) , 1 - 10 [ 2026-04-15 ] .

HAO Z , MU D , WANG L , et al . Intelligent Detection Method for Tunnel Leakage Water Based on Improved YOLOv10 [J/OL ] . Journal of South China University of Technology (Natural Science Edition) , 1 - 10 [ 2026-04-15 ] .

刘晓雅 , 许贵阳 , 白堂博 , 等 . 基于YOLO11-DFI的铁路沿线隐患物智能检测算法 [J/OL ] . 铁道建筑 , 1 - 9 [ 2026-01-14 ] .

LIU X , XU G , BAI T , et al . Intelligent Detection Algorithm for Hidden Hazards along Railway Lines Based on YOLO11-DFI [J/OL ] . Railway Construction , 1 - 9 [ 2026-01-14 ] .

张红瑞 , 冯威铭 , 杨潞霞 , 等 . 基于YOLO11改进的水下小目标检测算法CSAF-YOLO [J/OL ] . 计算机应用 , 1 - 12 [ 2026-01-14 ] .

ZHANG H , FENG W , YANG L , et al . The underwater small target detection algorithm CSAF-YOLO based on the improvement of YOLO11 [J/OL ] . Computer Applications , 1 - 12 [ 2026-01-14 ] .

张涛 , 田秀华 , 沈健 , 等 . 基于Rail-YOLO的铁路限界入侵人员检测算法研究 [J ] . 铁道通信信号 , 2026 , 62 ( 04 ): 34 - 42 .

ZHANG T , TIAN X , SHEN J , et al . Research on Railway Boundary Intrusion Personnel Detection Algorithm Based on Rail-YOLO [J ] . Railway Communication and Signal , 2026 , 62 ( 04 ): 34 - 42 .

LIN T , MAIRE M , BELONGIE J S , et al . Microsoft COCO: Common Objects in Context .[J ] . CoRR , 2014 , abs/1405. 0312

EVERINGHAM M , ESLAMI A M S , GOOL L , et al . The Pascal Visual Object Classes Challenge: A Retrospective [J ] . International Journal of Computer Vision , 2015 , 111 ( 1 ): 98 - 136 .

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

利用MPEG-4视频流的运动目标检测

一种高分辨率遥感图像视感知目标检测算法

一种改进的单步多框目标检测算法

基于YOLOv5s的轻量化朱鹮检测算法研究

从局部到全局的零参考低照度图像增强方法