双注意力引导的跨层优化交通场景语义分割

doi:10.16097/j.cnki.1009-6744.2023.01.025

交通运输系统工程与信息 ›› 2023, Vol. 23 ›› Issue (1): 236-244.DOI: 10.16097/j.cnki.1009-6744.2023.01.025

双注意力引导的跨层优化交通场景语义分割

谢新林^*1,2，罗臣彦^1,2，续欣莹³，谢刚^1,2,3

1.太原科技大学，电子信息工程学院，太原 030024；2.先进控制与装备智能化山西省重点实验室，太原 030024； 3.太原理工大学，电气与动力工程学院，太原 030024

收稿日期:2022-10-10 修回日期:2022-11-16 接受日期:2022-12-05 出版日期:2023-02-25 发布日期:2023-02-16
作者简介:谢新林(1990- )，男，山西运城人，讲师，博士。
基金资助:
国家自然科学基金(62006169)；山西省重点研发计划 (202102020101005)；太原科技大学博士科研启动基金(20192047)

Dual Attention Guided Cross-layer Optimized Traffic Scene Semantic Segmentation

XIE Xin-lin^*1,2, LUO Chen-yan^1,2, XU Xin-ying³, XIE Gang^1,2,3

1. School of Electronic and Information Engineering, Taiyuan University of Science and Technology, Taiyuan 030024, China; 2. Shanxi Key Laboratory of Advanced Control and Equipment Intelligence, Taiyuan 030024, China; 3. College of Electrical and Power Engineering, Taiyuan University of Technology, Taiyuan 030024, China

Received:2022-10-10 Revised:2022-11-16 Accepted:2022-12-05 Online:2023-02-25 Published:2023-02-16
Supported by:
National Natural Science Foundation of China (62006169)；Key Research and Development Program of Shanxi Province (202102020101005)；Taiyuan University of Science and Technology Scientific Research Initial Funding (20192047)

摘要/Abstract

摘要： 针对交通场景目标分割边缘不平滑以及小目标难以准确分割等问题，本文提出一种双注意力引导的跨层优化交通场景语义分割算法。首先，构建多分支特征提取编码网络，并利用串行非比例式空洞卷积实现空间上下文信息提取，进而改善小目标信息的丢失；其次，构建基于空间对齐的跨层特征融合解码网络，实现语义信息和细节信息的融合，增强不同尺度目标的表达能力；最后，提出通道和空间注意力机制，建模全局通道相关性和长距离位置相关性，增强网络对关键特征的学习能力。交通场景数据集Cityscapes和CamVid上的实验结果表明，所提特征提取编码网络、跨层特征融合解码网络以及注意力机制模块是有效的；所提语义分割算法获得了77.79%和78.66%的平均交并比，能够平滑目标分割边缘，尤其对细长条形目标具有鲁棒性。

关键词: 智能交通, 语义分割, 深度学习, 注意力机制, 小目标, 交通场景

Abstract: This paper proposes a dual attention guided cross-layer optimized traffic scene semantic segmentation to solve the problems that edge of object segmentation is not smooth and small objects are difficult to be accurately segmentedin the traffic scene. A coding network is established with multi-branch feature extraction. The assignment of dilation rate of atrous convolution is non-proportional and can extract spatial context information whichavoids the loss of small object information. Then, a cross-layer feature fusion decoding network based on spatial alignment is constructed to fuse semantic information and detail information, thereby enhancing the expressive ability of objects of different scales. At last, the channel and spatial attention mechanisms are proposed to model global channel correlations and long- distance location correlations, which enhances the ability to learn key features of the network. The experimental results on the traffic scene datasets Cityscapes and CamVid show that the proposed feature extraction encoding network, cross-layer feature fusion decoding network, and attention mechanism are effective.The proposed semantic segmentation algorithm achieves the mean intersection over union ratio of 77.79% and 78.66% , and can smooth the edge of object segmentation, especially for the long and thin objects.

Key words: intelligent transportation, semantic segmentation, deep learning, attention mechanism, small object, traffic scene

中图分类号:

U495
TP391.4

谢新林, 罗臣彦, 续欣莹, 谢刚. 双注意力引导的跨层优化交通场景语义分割[J]. 交通运输系统工程与信息, 2023, 23(1): 236-244.

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.

导出引用管理器 EndNote|Ris|BibTeX

链接本文: http://www.tseit.org.cn/CN/10.16097/j.cnki.1009-6744.2023.01.025

http://www.tseit.org.cn/CN/Y2023/V23/I1/236

参考文献

[1] 刘占文, 范颂华, 齐明远, 等. 基于时序融合的自动驾驶多任务感知算法[J]. 交通运输工程学报, 2021, 21(4): 223-234. [LIU Z W, FAN S H, QI M Y, et al. Multi-task perception algorithm of autonomous driving based on temporal fusion[J]. Journal of Traffic and Transportation Engineering, 2021, 21(4): 223-234.]

[2] ELHASSAN M A M, HUANG C, YANG C, et al. DSANet: Dilated spatial attention for real-time semantic segmentation in urban street scenes[J]. Expert Systems with Applications, 2021, 183: 115090.

[3] 罗会兰, 黎宵. 基于上下文和浅层空间编解码网络的图像语义分割方法[J]. 自动化学报, 2022, 48(7): 1834- 1846. [LUO H L, LI X. Image semantic segmentation method based on context and shallow space encoderdecoder network[J]. Acta Automatica Sinica, 2022, 48(7): 1834-1846.]

[4] HUYNH C, TRAN A T, LUU K, et al. Progressive semantic segmentation[C]. 2021 IEEE Conference on Computer Vision and Pattern Recognition. Los Alamitos: IEEE Computer Society Press, 2021: 16755-16764.

[5] LONG J, SHELHAMER E, DARRELL T. Fully convolutional networks for semantic segmentation[C]. 2015 IEEE Conference on Computer Vision and Pattern Recognition. Los Alamitos: IEEE Computer Society Press, 2015: 3431-3440.

[6] DONG G, YAN Y, SHEN C, et al. Real-time highperformance semantic image segmentation of urban street scenes[J]. IEEE Transactions on Intelligent Transportation Systems, 2020, 22(6): 3258-3274.

[7] 彭宏勤, 张国伍. 智慧交通与智慧物流:“交通 7+1 论坛”第五十五次会议纪实[J]. 交通运输系统工程与信息, 2019, 19(4): 1-4, 253. [PENG H Q, ZHANG G W. Intelligent transportation and intelligent logistics[J]. Journal of Transportation Systems Engineering and Information Technology, 2019, 19(4): 1-4, 253.]

[8] 高翔, 李春庚, 安居白. 基于注意力和多标签分类的图像实时语义分割[J]. 计算机辅助设计与图形学学报, 2021, 33(1): 59-67. [GAO X, LI C G, AN J B. Real-time image semantic segmentation based on attention mechanism and multi-label classification[J]. Journal of Computer-Aided Design & Computer Graphics, 2021, 33(1): 59-67.]

[9] 王瑞, 李霄峰, 史天运, 等. 基于视频深度学习的铁路周界入侵检测算法研究[J]. 交通运输系统工程与信息, 2020, 20(2): 61-68. [WANG R, LI X F, SHI T Y, et al. Railway perimeter intrusion detection algorithms based on video deep learning[J]. Journal of Transportation Systems Engineering and Information Technology, 2020, 20(2): 61-68.]

[10] CHEN L C, ZHU Y, PAPANDREOU G, et al. Encoderdecoder with atrous separable convolution for semantic image segmentation[C]. 2018 European Conference on Computer Vision. Heidelberg: Springer, 2018: 801-818.

[11] 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.]

[12] CHEN B, GONG C, YANG J. Importance-aware semantic segmentation for autonomous vehicles[J]. IEEE Transactions on Intelligent Transportation Systems, 2018, 20(1): 137-148.

[13] DING X, ZHANG X, MA N, et al. Repvgg: Making vggstyle convnets great again[C]. 2021 IEEE Conference on Computer Vision and Pattern Recognition. Los Alamitos: IEEE Computer Society Press, 2021: 13733-13742.

[14] PARK J, WOO S, LEE J Y, et al. Bam: Bottleneck attention module[C]. 2018 British Machine Vision Conference. Newcastle: British Machine Vision Association, 2018: 1-14.

[15] WOO S, PARK J, LEE J Y, et al. Cbam: Convolutional block attention module[C]. 2018 European Conference on Computer Vision. Heidelberg: Springer, 2018: 3-19.

[16] HOU Q, ZHOU D, FENG J. Coordinate attention for efficient mobile network design[C]. 2021 IEEE Conference on Computer Vision and Pattern Recognition. Nashville: IEEE Computer Society Press, 2021: 13713- 13722.

[17] HU X, JING L, SEHAR U. Joint pyramid attention network for real-time semantic segmentation of urban scenes[J]. Applied Intelligence, 2022, 52(1): 580-594.

双注意力引导的跨层优化交通场景语义分割

Dual Attention Guided Cross-layer Optimized Traffic Scene Semantic Segmentation

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

编辑推荐

Metrics

[1]	刘润坤, 于海洋, 程猛越, 任毅龙. 间歇性通信失效下网联自动驾驶队列控制策略[J]. 交通运输系统工程与信息, 2023, 23(2): 54-66.
[2]	张鹏, 顾云翔, 孙超, 李文权. 基于最小油耗的货车车速与信号配时协同优化模型[J]. 交通运输系统工程与信息, 2023, 23(2): 84-91.
[3]	施俊庆, 李睿, 程明慧, 阮俊辉, 谢星. 基于动态时空神经网络模型的地铁客流预测[J]. 交通运输系统工程与信息, 2023, 23(2): 139-147.
[4]	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.
[5]	唐进君, 付强, 王骋程, 袁琛, 张璇, 徐长靖. 高速公路可变限速控制策略多目标优化[J]. 交通运输系统工程与信息, 2023, 23(2): 252-261.
[6]	张冰涛, 常文文, 李秀兰. 基于时空脑电特征与并行神经网络的疲劳驾驶检测[J]. 交通运输系统工程与信息, 2023, 23(2): 315-325.
[7]	赵建东, 贺晓宇, 余智鑫, 韩明敏. 多网联范围下的智能网联车换道决策组合模型研究[J]. 交通运输系统工程与信息, 2023, 23(1): 77-85.
[8]	孙伟, 张梦雅, 马成元, 朱际宸, 杨晓光. 新型混合交通交叉口信号与车辆轨迹协同控制方法[J]. 交通运输系统工程与信息, 2023, 23(1): 97-105.
[9]	陈喜群, 朱奕璋, 吕朝锋. 基于混合近端策略优化的交叉口信号相位与配时优化方法[J]. 交通运输系统工程与信息, 2023, 23(1): 106-113.
[10]	贺文武, 裴博彧, 李雅婷, 刘小雨, 徐少兵. 基于双向自适应门控图卷积网络的交通流预测[J]. 交通运输系统工程与信息, 2023, 23(1): 187-197.
[11]	高雨虹, 曲昭伟, 宋现敏. 基于猎人猎物优化与双向长短时记忆组合模型的汽车出车率预测[J]. 交通运输系统工程与信息, 2023, 23(1): 198-206.
[12]	郝威, 龚雅馨, 张兆磊, 易可夫. 面向高速公路混合交通流的车辆协同合流策略[J]. 交通运输系统工程与信息, 2023, 23(1): 224-235.
[13]	杨柳, 杨莹, 宋允洲, 张宇. 驾驶压力影响因素与识别方法研究综述[J]. 交通运输系统工程与信息, 2022, 22(6): 40-50.
[14]	项新建, 胡海斌, 姚佳娜, 丁祎, 郑永平, 金立. 一种基于物影匹配算法的道路小目标跟踪方法[J]. 交通运输系统工程与信息, 2022, 22(6): 85-94.
[15]	陈志军, 胡军楠, 冷姚, 钱闯, 吴超仲. 基于轻量化网络和注意力机制的智能车快速目标识别方法[J]. 交通运输系统工程与信息, 2022, 22(6): 105-113.