Improved RT-DETR Framework for Railway Obstacle Detection

Obstacle intrusion detection in railway systems is a critical technology for ensuring the operational safety of trains. However, existing algorithms face challenges related to insufficient multiscale object detection, high model redundancy, and poor real-time performance. Building upon the RT-DETR f...

Full description

Saved in:
Bibliographic Details
Main Authors: Peng Li, Yanhui Peng, Su-Mei Wang, Cheng Zhong
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Access
Subjects:
Online Access:https://ieeexplore.ieee.org/document/11080059/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849420316173926400
author Peng Li
Yanhui Peng
Su-Mei Wang
Cheng Zhong
author_facet Peng Li
Yanhui Peng
Su-Mei Wang
Cheng Zhong
author_sort Peng Li
collection DOAJ
description Obstacle intrusion detection in railway systems is a critical technology for ensuring the operational safety of trains. However, existing algorithms face challenges related to insufficient multiscale object detection, high model redundancy, and poor real-time performance. Building upon the RT-DETR framework, this study proposes a Multiscale Separable Deformable (MSD) module that integrates depthwise convolution with deformable convolution to enhance feature extraction capabilities while reducing computational load. Additionally, a Deformable Agent Attention (DAA) mechanism is designed to optimize attention weights through sparse queries, effectively improving detection accuracy for small targets and enhancing inference speed in complex scenarios. Experimental results demonstrate that the improved model achieves 87.9% mean average precision (mAP) on a railway dataset, with a detection speed of 90 frames per second (FPS). The proposed model achieves a +1.7% mAP improvement and 13.9% faster inference speed compared to RT-DETR, while simultaneously reducing model parameters by 24.6%. As a result, the proposed model is highly effective for multiple obstacle intrusion detection in complex real-world scenarios.
format Article
id doaj-art-a30fdd0b100a43588bed62797ea1bb2e
institution Kabale University
issn 2169-3536
language English
publishDate 2025-01-01
publisher IEEE
record_format Article
series IEEE Access
spelling doaj-art-a30fdd0b100a43588bed62797ea1bb2e2025-08-20T03:31:47ZengIEEEIEEE Access2169-35362025-01-011312586912588010.1109/ACCESS.2025.358915911080059Improved RT-DETR Framework for Railway Obstacle DetectionPeng Li0https://orcid.org/0000-0001-9558-5088Yanhui Peng1Su-Mei Wang2https://orcid.org/0000-0002-1285-1553Cheng Zhong3School of Railway Tracks and Transportation, Wuyi University, Jiangmen, ChinaSchool of Railway Tracks and Transportation, Wuyi University, Jiangmen, ChinaNational Rail Transit Electrification and Automation Engineering Technology Research Center (Hong Kong Branch), Guangzhou, Hong KongSchool of Railway Tracks and Transportation, Wuyi University, Jiangmen, ChinaObstacle intrusion detection in railway systems is a critical technology for ensuring the operational safety of trains. However, existing algorithms face challenges related to insufficient multiscale object detection, high model redundancy, and poor real-time performance. Building upon the RT-DETR framework, this study proposes a Multiscale Separable Deformable (MSD) module that integrates depthwise convolution with deformable convolution to enhance feature extraction capabilities while reducing computational load. Additionally, a Deformable Agent Attention (DAA) mechanism is designed to optimize attention weights through sparse queries, effectively improving detection accuracy for small targets and enhancing inference speed in complex scenarios. Experimental results demonstrate that the improved model achieves 87.9% mean average precision (mAP) on a railway dataset, with a detection speed of 90 frames per second (FPS). The proposed model achieves a +1.7% mAP improvement and 13.9% faster inference speed compared to RT-DETR, while simultaneously reducing model parameters by 24.6%. As a result, the proposed model is highly effective for multiple obstacle intrusion detection in complex real-world scenarios.https://ieeexplore.ieee.org/document/11080059/Convolutional neural network (CNN)deep learningobstacle intrusion detectionrailway traffictransformer
spellingShingle Peng Li
Yanhui Peng
Su-Mei Wang
Cheng Zhong
Improved RT-DETR Framework for Railway Obstacle Detection
IEEE Access
Convolutional neural network (CNN)
deep learning
obstacle intrusion detection
railway traffic
transformer
title Improved RT-DETR Framework for Railway Obstacle Detection
title_full Improved RT-DETR Framework for Railway Obstacle Detection
title_fullStr Improved RT-DETR Framework for Railway Obstacle Detection
title_full_unstemmed Improved RT-DETR Framework for Railway Obstacle Detection
title_short Improved RT-DETR Framework for Railway Obstacle Detection
title_sort improved rt detr framework for railway obstacle detection
topic Convolutional neural network (CNN)
deep learning
obstacle intrusion detection
railway traffic
transformer
url https://ieeexplore.ieee.org/document/11080059/
work_keys_str_mv AT pengli improvedrtdetrframeworkforrailwayobstacledetection
AT yanhuipeng improvedrtdetrframeworkforrailwayobstacledetection
AT sumeiwang improvedrtdetrframeworkforrailwayobstacledetection
AT chengzhong improvedrtdetrframeworkforrailwayobstacledetection