基于加权密集连接卷积网络的快速交通标志检测

交通运输系统工程与信息 ›› 2020, Vol. 20 ›› Issue (2): 48-54.

• 智能交通系统与信息技术 • 上一篇下一篇

基于加权密集连接卷积网络的快速交通标志检测

邵毅明，屈治华，邓天民^*，宋晓华

重庆交通大学交通运输学院，重庆 400074

收稿日期:2019-09-12 修回日期:2019-10-24 出版日期:2020-04-25 发布日期:2020-04-30
作者简介:邵毅明(1955-)，男，四川资阳人，教授，博士.
基金资助:
国家重点研发计划/ National Key Research and Development Program of China(2016YFB0100905)；重庆市科技人才培养计划/ Chongqing Science and Technology Talent Training Program(cstc2013kjrc-qnrc0148).

Fast Traffic Sign Detection Based on Weighted Densely Connected Convolutional Network

SHAO Yi-ming, QU Zhi-hua, DENG Tian-min, SONG Xiao-hua

School of Traffic and Transportation, Chongqing Jiaotong University, Chongqing 400074, China

Received:2019-09-12 Revised:2019-10-24 Online:2020-04-25 Published:2020-04-30

摘要/Abstract

摘要：

为提高算法对交通标志快速定位的准确性，改善现有检测算法在复杂交通环境下检测效果不佳、实时性较差的问题，提出一种基于动态加权密集连接卷积网络的交通标志快速检测算法. 选用YOLOv2 作为基础网络，通过增加动态加权密集块对各层特征图的权重进行调节，实现深层高语义信息和浅层低语义信息的融合；使用MobileNet 轻量化网络结构，通过可分离卷积操作有效降低网络的计算成本；针对池化操作中图像特征丢失严重问题引入 CBAM模块，利用通道注意力和空间注意力信息增强关键特征的表达能力. 实验结果表明，本文算法在GTSDB数据集上分别达到了96.14%的检测精度和139 frame/s 的检测速度，在保证较高检测精度的同时，能够有效提高检测效率，满足实时检测要求.

关键词: 智能交通, 交通标志检测, 密集连接网络, 深度学习, MobileNet

Abstract:

The current traffic sign detection algorithms appear to have poor detection effect and low practicability under complex traffic environments. This paper proposed a fast traffic sign detection algorithm based on dynamic weighted densely connected convolutional network. The purpose is to enhance the detection accuracy of the algorithm and identify traffic signs with improved speed. YOLOv2 was selected as the basic network, and the weight of each layer's feature map was adjusted by adding dynamic weighted densely blocks. The deep highsemantic information and shallow low-semantic information was integrated subsequently. The lightweight network structure in MobileNet and the separable convolution operation reduced the network computing costs effectively. To resolve the image feature loss problem in pooling operation, the study used the Convolutional Block Attention Module (CBAM) to enhance the performance of key features by using channel attention and spatial attention information. The experimental results confirmed that the proposed algorithm achieved a detection accuracy of 96.14%, and the detection speed was 139 frame/s on the German Traffic Sign Detection Benchmark (GTSDB) dataset. The algorithm effectively improved the detection efficiency and practicability, and maintained good detection accuracy.

Key words: intelligent transportation, traffic sign detection, densely connected network, deep learning, MobilNet

中图分类号:

TP391.41

邵毅明，屈治华，邓天民，宋晓华. 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 Yi-ming, QU Zhi-hua, DENG Tian-min, SONG Xiao-hua. 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.

导出引用管理器 EndNote|Ris|BibTeX

链接本文: http://www.tseit.org.cn/CN/abstract/abstract20009.shtml

http://www.tseit.org.cn/CN/Y2020/V20/I2/48

参考文献

[1] 张新钰, 高洪波, 赵建辉, 等. 基于深度学习的自动驾驶技术综述[J]. 清华大学学报(自然科学版), 2018, 58 (4): 438-444. [ZHANG X Y, GAO H B, ZHAO J H, et al. Overview of deep learning intelligent driving methods[J]. Journal of Tsinghua University (Science and Technology), 2018, 58(4): 438-444.]

[2] 刘华平, 李建民, 胡晓林, 等. 动态场景下的交通标识检测与识别研究进展[J]. 中国图像图形学报, 2013, 18 (5): 493-503. [LIU H P, LI J M, HU X L, et al. Recent progress in detection and recognition of the traffic signs in dynamic scenes[J]. Journal of Image and Graphics, 2013, 18(5): 493-503.]

[3] 张秀玲, 张逞逞, 周凯旋. 基于感兴趣区域的CNNSqueeze 交通标志图像识别[J]. 交通运输系统工程与信息, 2019, 19(3): 48-53. [ZHANG X L, ZHANG C C, ZHOU K X. Traffic sign image recognition via CNNSqueeze based on region of interest[J]. Journal of Transportation Systems Engineering and Information Technology, 2019, 19(3): 48-53.]

[4] 黄继鹏, 史颖欢, 高阳. 面向小目标的多尺度Faster- RCNN 检测算法[J]. 计算机研究与发展, 2019, 56(2): 319-327. [HAUNG J P, SHI Y H, GAO Y. Multi-scale Faster-RCNN algorithm for object detection[J]. Journal of Computer Research and Development, 2019, 56(2): 319-327.]

[5] GIRSHICK R, DONAHUE J, DARRELL T, et al. Region- based convolutional networks for accurate object detection and segmentation[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2015, 38 (1): 142-158.

[6] REDMON J, FARHADI A. YOLO9000: Better, faster, stronger[C]. IEEE Conference on Computer Vision and Pattern Recognition. IEEE, 2017: 6517-6525.

[7] LIU W, ANGUELOV D, ERHAN D, et al. SSD: Single shot multibox detector[C]. European Conference on Computer Vision. Springer, Cham, 2016: 21-37.

[8] 薛玉利. 雾霾天气情况下的交通标志检测[J]. 交通运输系统工程与信息, 2016, 16(4): 88-94. [XUE Y L. Traffic sign detection under fog and haze weather[J], Journal of Transportation Systems Engineering and Information Technology, 2016, 16(4): 88-94.]

[9] MOGELMOSE A, TRIVEDI M M, MOESLUND T B. Vision-based traffic sign detection and analysis for intelligent driver assistance systems: perspectives and survey[J]. IEEE Transactions on Intelligent Transportation Systems, 2012, 13(4): 1484-1497.

[10] HUANG G, LIU Z, WEINBERGER K Q, et al. Densely connected convolutional networks[C]. IEEE Conference on Computer Vision and Pattern Recognition, 2016: 4700-4708.

[11] HOWARD A G, ZHU M, CHEN B, et al. MobileNets: Efficient convolutional neural networks for mobile vision applications[OL]. [2019-09-03]. http://arxiv.orb/abs/ 1704.04861.

[12] 李珣, 刘瑶, 李鹏飞, 等. 基于Darknet框架下YOLO v2 算法的车辆多目标检测方法[J].交通运输工程学报, 2018, 18(6): 142- 158. [LI X, LIU Y, LI P F, et al. Vehicle multi-target detection method based on YOLO v2 algorithm under darknet framework[J]. Journal of Traffic and Transportation Engineering, 2018, 18(6): 142-158.]

[13] WOO S , PARK J , LEE J Y , et al. CBAM: convolutional block attention module[C]. European Conference on Computer Vision. Springer, 2018.

[14] HOUBEN S, STALLKAMP J, SALMEN J, et al. Detection of traffic signs in real-world images: The german traffic sign detection benchmark[C]. IEEE International Joint Conference on Neural Networks, 2013: 1-8.

[15] LOKANATH M, KUMAR K S, KEERTHI E S. Accurate object classification and detection by faster- RCNN[J]. IOP Conference Series: Materials Science and Engineering, 2017, 263: 052028.

[16] VIOLA P, JONES M. Rapid object detection using a boosted cascade of simple features[C]. IEEE Conference on Computer Vision and Pattern Recognition, 2001: 511-518.

[17] ZHANG X Y, ZHOU X Y, LIU M X, et al. ShuffleNet: an extremely efficient convolutional neural network for mobile devices[OL]. [2019-09-03]. http://arxiv.orb/abs/ 1707.01083.

[18] IANDOLA F N, HAN S, MOSKEWICZ M W, et al. Squeeze Net: Alex Net-level accuracy with 50x fewer parameters and <0.5MB model size[C]. International Conference on Learning Representations, 2016.

基于加权密集连接卷积网络的快速交通标志检测

Fast Traffic Sign Detection Based on Weighted Densely Connected Convolutional Network

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]	郭烈, 胥林立, 秦增科, 王旭. 自动驾驶接管影响因素分析与研究进展[J]. 交通运输系统工程与信息, 2022, 22(2): 72-90.
[2]	徐猛, 刘涛, 钟绍鹏, 姜宇. 城市智慧公交研究综述与展望[J]. 交通运输系统工程与信息, 2022, 22(2): 91-108.
[3]	程君, 张立炎, 陈启宏. 一种基于地图辅助的自动驾驶视-惯融合定位方法[J]. 交通运输系统工程与信息, 2022, 22(2): 117-126.
[4]	王国栋, 刘立, 孟宇, 马智萍, 郑淏清, 顾青, 白国星. 一体式车辆避撞轨迹规划与跟踪控制[J]. 交通运输系统工程与信息, 2022, 22(2): 127-136.
[5]	马东方, 陈曦, 吴晓东, 金盛. 基于强化学习的干线信号混合协同优化方法[J]. 交通运输系统工程与信息, 2022, 22(2): 145-153.
[6]	许波桅, 王玲玲, 李军军. 自动化码头多级作业超网络延误传导特性研究[J]. 交通运输系统工程与信息, 2022, 22(2): 280-289.
[7]	田会娟, 刘嘉伟, 翟佳豪, 邓琳琳. 基于多入侵线的视频车速检测方法[J]. 交通运输系统工程与信息, 2022, 22(1): 49-56.
[8]	贾彦峰, 曲大义, 赵梓旭, 王韬, 宋慧. 基于安全势场的网联自主车辆跟驰行为决策及模型[J]. 交通运输系统工程与信息, 2022, 22(1): 85-97.
[9]	刘文, 汪文博. 基于秩最小化矩阵去噪的船舶轨迹重构方法[J]. 交通运输系统工程与信息, 2022, 22(1): 106-114.
[10]	赵建东, 陈溱, 焦彦利, 张凯丽, 韩明敏. 重点营运车辆的异常驾驶行为识别研究[J]. 交通运输系统工程与信息, 2022, 22(1): 282-291.
[11]	徐东伟, 彭航, 商学天, 魏臣臣, 杨艳芳. 基于图自编码-生成对抗网络的路网数据修复[J]. 交通运输系统工程与信息, 2021, 21(6): 33-41.
[12]	游峰, 梁健中, 曹水金, 肖智豪, 吴镇江, 王海玮. 面向多目标跟踪的密集行人群轨迹提取和运动语义感知[J]. 交通运输系统工程与信息, 2021, 21(6): 42-54.
[13]	王福建, 俞佳浩, 赵锦焕, 梅振宇. 基于站点实时关联度的短时公交客流预测方法[J]. 交通运输系统工程与信息, 2021, 21(6): 131-144.
[14]	张毅, 姚丹亚, 李力, 裴华鑫, 晏松, 葛经纬. 智能车路协同系统关键技术与应用[J]. 交通运输系统工程与信息, 2021, 21(5): 40-51.
[15]	贺正冰, 徐瑞康, 谢东繁, 宗芳, 钟任新. 数据驱动跟驰模型综述[J]. 交通运输系统工程与信息, 2021, 21(5): 102-113.