Semantic Segmentation of Railway Scene Based on Reticulated
Multi-scale and Bidirectional Channel Attention

doi:10.16097/j.cnki.1009-6744.2023.02.025

Abstract

Abstract: Semantic segmentation is the basis of intelligent perception. This paper proposes a semantic segmentation network for railway scenes based on reticulated multi-scale fusion and bidirectional channel attention to address the difficulties that railway scene categories are complex and effective features are challenging to extract. To enhance the discriminant ability of the model for various railway facilities, a reticulated multi-scale fusion module is proposed. The module is embedded in the backbone network to obtain the parallel connection of features of different scales and carries out reticulated information interaction in the fusion layer to realize the feature fusion of different branches. By aggregating inputs from other branches, the output of the model can simultaneously retain multi-resolution features. In order to improve the extraction performance of effective features in complex railway scenes, a bidirectional channel attention module is proposed. After the up-down sampling operation of the backbone network, the forward channel attention module makes the output feature map weighted by the input features of different scales, so as to adaptively improve the expression of effective features. The inverted channel attention module is inserted before the final output of the model, to generate effective high-level semantic information while retaining the underlying spatial information. Experimental results on the RailSem19 railway dataset show that the proposed method can significantly improve the segmentation performance of easily confused categories as well as railway facility categories such as track area, catenary mast, train and protective fence, and mIoU reached 65.12%, which is a certain improvement compared with other methods.

Key words: intelligent transportation, railway semantic segmentation, deep learning, complex railway scene, reticulated multi-scale fusion, bidirectional channel attention

摘要： 语义分割是智能感知的基础。本文针对复杂铁路场景类别易混淆及有效特征难提取的问题，提出一种基于网状多尺度融合与双向通道注意力的铁路场景语义分割网络。为增强模型对各类铁路设施的判别能力，提出网状多尺度融合模块。该模块嵌入主干网络，获取不同尺度的特征并行连接，并在融合层进行网状信息交互，实现不同分支的特征融合。通过汇聚其他分支输入，模型输出可同时保留多分辨率特征。为提升复杂铁路场景下有效特征的提取性能，提出双向通道注意力模块。正向通道注意力模块位于主干网络上下采样操作后，使输出特征图由不同尺度的输入特征加权重组，从而自适应地提升有效特征的表达；反向通道注意力模块插入模型最终输出之前，保留底层空间信息的同时，生成有效高阶语义信息。在RailSem19铁路数据集上的实验结果表明，本文所述方法对于易混淆类别，以及轨道区域、接触网立柱、列车及防护栅栏等铁路设施类别的分割性能均有显著提升，mIoU达到65.12%，相比于其他方法有一定提升。

关键词: 智能交通, 铁路语义分割, 深度学习, 复杂铁路场景, 网状多尺度融合, 双向通道注意力

CLC Number:

U495

LU Tong, YU Zu-jun, GUO Bao-qing, RUAN Tao. Semantic Segmentation of Railway Scene Based on Reticulated Multi-scale and Bidirectional Channel Attention[J]. Journal of Transportation Systems Engineering and Information Technology, 2023, 23(2): 233-241.

路通, 余祖俊, 郭保青, 阮涛. 网状多尺度与双向通道注意力的铁路场景语义分割[J]. 交通运输系统工程与信息, 2023, 23(2): 233-241.

Add to citation manager EndNote|Ris|BibTeX

URL: http://www.tseit.org.cn/EN/10.16097/j.cnki.1009-6744.2023.02.025

http://www.tseit.org.cn/EN/Y2023/V23/I2/233

References

[1] 李兴鑫, 朱力强, 余祖俊. 自适应铁路场景前景目标检测[J]. 交通运输系统工程与信息, 2020, 20(2): 83-90. [LI X X, ZHU L Q, YU Z J. Adaptive foreground object detection in railway sence[J]. Journal of Transportation Systems Engineering and Information Technology, 2020, 20(2): 83-90.]

[2] 郭保青, 王宁. 基于改进深度卷积网络的铁路入侵行人分类算法[J]. 光学精密工程, 2018, 26(12): 3040- 3050. [GUO B Q, WANG N. Pedestrian intruding railway clearance classification algorithm based on improved deep convolutional network[J]. Optics and Precision Engineering, 2018, 26(12): 3040-3050.]

[3] 邵毅明, 屈治华, 邓天民, 等. 基于加权密集连接卷积网络的快速交通标志检测[J]. 交通运输系统工程与信息, 2020, 20(2): 48-54. [SHAO Y M, QU Z H, DENG T M, et al. Fast traffic sign detection based on weighted densely connected convolutional network[J]. Journal of Transportation Systems Engineering and Information Technology, 2020, 20(2): 48-54.]

[4] 熊敏君, 李晨, 张慧源, 等. 基于嵌入式GPU平台的列车运行环境检测算法[J]. 控制与信息技术, 2021(4): 72- 77. [XIONG M J, LI C, ZHANG H Y, et al. Train running environment detection algorithm based on embedded GPU platform[J]. Control and Information Technology, 2021(4): 72-77.]

[5] WANG Y, WANG L D, HEN H Y, et al. Railnet: A segmentation network for railroad detection[J]. IEEE Access, 2019(99): 1.

[6] 王洋, 朱力强, 余祖俊, 等. 高速铁路场景的分割与识别算法[J]. 光学学报, 2019, 39(6): 119-126. [WANG Y, ZHU L Q, YU Z J, et al. Segmentation and recognition algorithm for high-speed railway scene[J]. Acta Optica Sinica, 2019, 39(6): 119-126.]

[7] 赵冰, 李平, 代明睿, 等. 基于深度学习的铁路图像场景分类优化研究[J]. 计算机系统应用, 2019, 28(6): 228-234. [ZHAO B, LI P, DAI M R, et al. Research on optimization method of railway image scene classification based on deep learning method[J]. Computer Systems & Applications, 2019, 28(6): 228-234.]

[8] RONNEBERGER O, FISCHER P, BROX T. U-net: Convolutional networks for biomedical image segmentation[J]. Proceedings of the 18th International Conference on Medical Image Computing and ComputerAssisted Intervention, Munich: Springer, 2015: 234-241.

[9] SUN K, XIAO B, LIU D, et al. Deep high-resolution representation learning for human pose estimation[C]// 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Long Beach, CA, USA: IEEE, 2019.

[10] ZHAO H, SHI J, QI X, et al. Pyramid scene parsing network[C]// 2017 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Honlulu, HI, USA: IEEE, 2017.

[11] HE K, ZHANG X, REN S, et al. Deep residual learning for image recognition[C]//2016 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, USA: IEEE, 2016.

[12] 孙剑明, 韩生权, 沈子成, 等. 基于双卷积链的双目人体姿态距离定位识别[J]. 兵工学报, 2022, 43(11): 2846-2854. [SUN J M, HAN S Q, SHEN Z C, et al. Binocular human pose and distance identification based on double convolutional chain[J]. Acta Armamentarii, 2022, 43(11): 2846-2854.]

[13] 尉天成, 陈小锋, 殷元亮. 基于多尺度卷积神经网络的道路交通标志识别方法研究[J]. 西北工业大学学报, 2021, 39(4): 891-900. [WEI T C, CHEN X F, YIN Y L. Research on traffic sign recognition method based on multi-scale convolution neural network[J]. Journal of Northwestern Polytechnical University, 2021, 39(4): 891- 900.]

[14] WOO S, PARK J, LEE J Y, et al. CBAM: Convolutional block attention module[C]. Springer: European Conference on Computer Vision (ECCV), 2018.

[15] SANDLER M, HOWARD A, ZHU M, et al. MobileNetV2: Inverted residuals and linear bottlenecks [C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Salt Lake City, UT, USA: IEEE, 2018.

[16] ZENDEL O, MURSCHITZ M, ZEILINGER M, et al. Railsem19: A dataset for semantic rail scene understanding[C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Long Beach, CA, USA, IEEE, 2019.

[17] ZHANG H, DANA K, SHI J, et al. Context encoding for semantic segmentation[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Salt Lake City, UT, USA, IEEE, 2018.

[18] KIRILLOV A, GIRSHICK R, HE K, et al. Panoptic feature pyramid networks[C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Long Beach, CA, USA, IEEE, 2019.

[19] FU J, LIU J, TIAN H, et al. Dual attention network for scene segmentation[C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Long Beach, CA, USA, IEEE, 2019.

[20] KIRILLOV A, WU Y, HE K, et al. PointRend: Image segmentation as rendering[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Seattle, WA, USA, IEEE, 2020.

[21] YIN M H, YAO Z L, CAO Y, et al. Disentangled nonlocal neural networks[C]. Glasgow UK: European Conference on Computer Vision (ECCV), 2020.

Semantic Segmentation of Railway Scene Based on Reticulated Multi-scale and Bidirectional Channel Attention

网状多尺度与双向通道注意力的铁路场景语义分割

PDF

Knowledge

Abstract

Cite this article

share this article

References

Related Articles 15

Recommended Articles

Metrics

[1]	GUO Yu-dong, YANG Fei, ZHOU Tao, YAO Zhen-xing, ZHANG Chu-liang, WEI Yin-cheng. Two-layer Model to Distinguish Urban Motorized Travel Mode Based on Mobile Phone Signaling Data [J]. Journal of Transportation Systems Engineering and Information Technology, 2023, 23(3): 101-109.
[2]	HAN Lei, ZHANG Lun, GUO Wei-an. Variable Speed Limit Control Based on Improved Dueling Double Deep Q Network Under Mixed Traffic Environment [J]. Journal of Transportation Systems Engineering and Information Technology, 2023, 23(3): 110-122.
[3]	LIU Run-kun, YU Hai-yang, CHENG Meng-yue, REN Yi-long. Platoon Control Strategy for Connected and Automated Vehicles Under Intermittent Communication Failures [J]. Journal of Transportation Systems Engineering and Information Technology, 2023, 23(2): 54-66.
[4]	ZHANG Peng, GU Yun-xiang, SUN Chao, LI Wen-quan. Collaborative Optimization Model of Truck Speed and Signal Timing Based on Minimum Fuel Consumption [J]. Journal of Transportation Systems Engineering and Information Technology, 2023, 23(2): 84-91.
[5]	TANG Jin-jun, FU Qiang, WANG Cheng-cheng, YUAN Chen, ZHANG Xuan, XU Chang-jing. Multi-objective Optimization of Variable Speed Limit Control Strategy on Expressway [J]. Journal of Transportation Systems Engineering and Information Technology, 2023, 23(2): 252-261.
[6]	ZHANG Bing-tao, CHANG Wen-wen, LI Xiu-lan. Fatigue Driving Detection Based on Spatial-temporal Electroencephalogram Features and Parallel Neural Networks [J]. Journal of Transportation Systems Engineering and Information Technology, 2023, 23(2): 315-325.
[7]	ZHAO Jian-dong, HE Xiao-yu, YU Zhi-xin, HAN Ming-min. A Combination Model for Connected and Autonomous Vehicles Lane-changing Decision-making Under Multi Connectivity Range [J]. Journal of Transportation Systems Engineering and Information Technology, 2023, 23(1): 77-85.
[8]	SUN Wei, ZHANG Meng-ya, MA Cheng-yuan, ZHU Ji-chen, YANG Xiao-guang. Coordination of Signal and Vehicle Trajectory at Intersections for Mixed Traffic Flow [J]. Journal of Transportation Systems Engineering and Information Technology, 2023, 23(1): 97-105.
[9]	CHEN Xi-qun, ZHU Yi-zhang, LV Chao-feng. Signal Phase and Timing Optimization Method for Intersection Based on Hybrid Proximal Policy Optimization [J]. Journal of Transportation Systems Engineering and Information Technology, 2023, 23(1): 106-113.
[10]	HE Wen-wu, PEI Bo-yu, LI Ya-ting, LIU Xiao-yu, XU Shao-bing. Traffic Flow Forecasting Based on Bi-directional Adaptive Gating Graph Convolutional Networks [J]. Journal of Transportation Systems Engineering and Information Technology, 2023, 23(1): 187-197.
[11]	GAO Yu-hong, , QU Zhao-wei , SONG Xian-min . Car Operation Rate Prediction Based on Combination Model of Hunter-prey Optimizer Algorithm and Bi-directional Long Short-term Memory Neural Network [J]. Journal of Transportation Systems Engineering and Information Technology, 2023, 23(1): 198-206.
[12]	HAO Wei, GONG Ya-xin, ZHANG Zhao-lei, YI Ke-fu. Cooperative Merging Strategy for Freeway Ramp in a Mixed Traffic Environment [J]. Journal of Transportation Systems Engineering and Information Technology, 2023, 23(1): 224-235.
[13]	XIE Xin-lin, LUO Chen-yan, XU Xin-ying, XIE Gang. Dual Attention Guided Cross-layer Optimized Traffic Scene Semantic Segmentation [J]. Journal of Transportation Systems Engineering and Information Technology, 2023, 23(1): 236-244.
[14]	YANG Liu, YANG Ying, SONG Yun-zhou, ZHANG Yu. A Review of Influencing Factors and Identification Methods of Driver Stress [J]. Journal of Transportation Systems Engineering and Information Technology, 2022, 22(6): 40-50.
[15]	XIANG Xin-jian, HU Hai-bin, YAO Jia-na, DING Yi, ZHENG Yong-ping, JIN Li. A Road Small Target Tracking Approach Based on Object-shadow Matching Algorithm [J]. Journal of Transportation Systems Engineering and Information Technology, 2022, 22(6): 85-94.