YOLOv3-A：基于注意力机制的交通标志检测网络

doi:10.11959/j.issn.1000-436x.2021031

通信学报 ›› 2021, Vol. 42 ›› Issue (1): 87-99.doi: 10.11959/j.issn.1000-436x.2021031

YOLOv3-A：基于注意力机制的交通标志检测网络

郭璠, 张泳祥, 唐琎, 李伟清

中南大学自动化学院，湖南长沙 410083

修回日期:2020-09-28 出版日期:2021-01-25 发布日期:2021-01-01
作者简介:郭璠（1982- ），女，湖南临澧人，博士，中南大学副教授、硕士生导师，主要研究方向为图像处理、计算机视觉、人工智能等。
张泳祥（1994- ），男，河南安阳人，中南大学硕士生，主要研究方向为模式识别、图像处理等。
唐琎（1966- ），男，湖南武冈人，博士，中南大学教授、博士生导师，主要研究方向为计算机视觉、机器人、嵌入式系统、智能信息处理等。
李伟清（1997- ），男，河南信阳人，中南大学硕士生，主要研究方向为医学图像处理、机器学习等。
基金资助:
国家自然科学基金资助项目(61502537);湖南省自然科学基金资助项目(2018JJ3681);中南大学中央高校基本科研业务费专项基金资助项目(2020zzts567)

YOLOv3-A: a traffic sign detection network based on attention mechanism

Fan GUO, Yongxiang ZHANG, Jin TANG, Weiqing LI

School of Automation, Central South University, Changsha 410083, China

Revised:2020-09-28 Online:2021-01-25 Published:2021-01-01
Supported by:
The National Natural Science Foundation of China(61502537);The Natural Science Foundation of Hunan Province(2018JJ3681);The Fundamental Research Funds for the Central Universities of Central South University(2020zzts567)

摘要/Abstract

摘要：

为了解决已有YOLOv3算法对于存在小目标问题和背景复杂问题的交通标志检测任务会有较多的误检和漏检的问题，在YOLOv3算法的基础上，提出了目标检测的通道注意力方法和基于语义分割引导的空间注意力方法，形成 YOLOv3-A 算法。YOLOv3-A 算法通过对检测分支特征在通道和空间 2 个维度进行重新标定，使网络聚焦和增强有效特征，并抑制干扰特征，提高了算法的检测能力。在TT100K交通标志数据集上的实验表明，所提算法对小目标检测性能的改善尤为明显，相比于YOLOv3算法，所提算法的精度和召回率分别提升了1.9%和2.8%。

关键词: 交通标志检测, 小目标检测, 注意力机制, 语义分割

Abstract:

To solve the problem that the existing YOLOv3 algorithm had more false detections and missed detections for traffic sign detection task with small target problems and complex background, based on the YOLOv3, a channel attention method for target detection and a spatial attention method based on semantic segmentation guidance were proposed to form the YOLOv3-A (attention) algorithm.The detection features in the channel and spatial dimensions were recalibrated, allowing the network to focus and enhance the effective features, and suppress interference features, which greatly improved the detection performance.Experiments on the TT100K traffic sign data set show that the algorithm improves the detection performance of small targets, and the accuracy and recall rate of the YOLOv3 are improved by 1.9% and 2.8% respectively.

Key words: traffic sign detection, small target detection, attention mechanism, semantic segmentation

中图分类号:

TP391.41

郭璠, 张泳祥, 唐琎, 李伟清. YOLOv3-A：基于注意力机制的交通标志检测网络[J]. 通信学报, 2021, 42(1): 87-99.

Fan GUO, Yongxiang ZHANG, Jin TANG, Weiqing LI. YOLOv3-A: a traffic sign detection network based on attention mechanism[J]. Journal on Communications, 2021, 42(1): 87-99.

图/表 18

图1

图2

图3

图4

图5

图6

图7

图8

图9

图10

图11

图12

表1

4种特征压缩方法对比"

特征压缩方法	Precision	Recall
GMaxPool	86.9%	89.8%
GAvgPool	86.7%	89.6%
GMaxPool + GAvgPool	87.0%	90.1%
Concat(GmaxPool, GAvgPool)	$87 . 2 %$	$90 . 3 %$

表1

表2

图13

图14

图15

表3

4种检测方法的AUC对比"

方法	小目标	中目标	大目标	整体尺度目标
文献[25]方法	0.854	0.921	0.915	0.897
RetinaNet101^[8]	0.891	0.957	0.873	0.920
YOLOv3^[14]	0.876	0.952	0.914	0.917
YOLOv3-A	$0 . 911$	$0 . 965$	$0 . 938$	$0 . 941$

表3

参考文献 28

[1]	DALAL N , TRIGGS B . Histograms of oriented gradients for human detection[C]// 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2005: 886-893.
[2]	LEE T S . Image representation using 2D Gabor wavelets[J]. IEEE Transactions on Pattern Analysis ＆ Machine Intelligence, 1996,18(10): 959-971.
[3]	VIOLA P A , JONES M J . Rapid object detection using a boosted cascade of simple features[C]// IEEE Computer Society Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2001: 511-518.
[4]	REN S , HE K , GIRSHICK R ,et al. Faster R-CNN:towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017,39(6): 1137-1149.
[5]	REDMON J , DIVVALA S , GIRSHICK R ,et al. You only look once:Unified,real-time object detection[C]// 2016 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2016: 779-788.
[6]	LIU W , ANGUELOV D , ERHAN D ,et al. SSD:single shot multibox detector[C]// European Conference on Computer Vision. Berlin:Springer, 2016: 21-37.
[7]	RAJENDRAN S P , SHINE L , PRADEEP R ,et al. Fast and accurate traffic sign recognition for self driving cars using RetinaNet based detector[C]// 2019 International Conference on Communication and Electronics Systems. Piscataway:IEEE Press, 2019: 784-790.
[8]	LIN T Y , GOYAL P , GIRSHICK R ,et al. Focal loss for dense object detection[C]// 2017 IEEE International Conference on Computer Vision. Piscataway:IEEE Press, 2017: 2980-2988.
[9]	HE K , ZHANG X , REN S ,et al. Deep residual learning for image recognition[C]// 2016 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2016: 770-778.
[10]	HOUBEN S , STALLKAMP J , SALMEN J ,et al. Detection of traffic signs in real-world images:the German traffic sign detection benchmark[C]// The 2013 International Joint Conference on Neural Networks. Piscataway:IEEE Press, 2013: 1-8.
[11]	YANG Y , LIU S , MA W ,et al. Efficient traffic-sign recognition with scale-aware CNN[C]// British Machine Vision Conference. London:BMVA Press, 2017: 1-13.
[12]	LARSSON F , FELSBERG M . Using Fourier descriptors and spatial models for traffic sign recognition[C]// 2011 Scandinavian Conference on Image Analysis(SCIA 2011). Berlin:Springer, 2011: 238-249.
[13]	MENG Z , FAN X , CHEN X ,et al. Detecting small signs from large images[C]// 2017 IEEE International Conference on Information Reuse and Integration. Piscataway:IEEE Press, 2017: 217-224.
[14]	REDMON J , FARHADI A . YOLOv3:an incremental improvement[J]. arXiv Preprint,arXiv:1804.02767, 2018.
[15]	LIN T Y , DOLLáR P , GIRSHICK R ,et al. Feature pyramid networks for object detection[C]// 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2017: 2117-2125.
[16]	HU J , SHEN L , SUN G . Squeeze-and-excitation networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018,42(8): 2011-2023.
[17]	WANG F , JIANG M , QIAN C ,et al. Residual attention network for image classification[C]// 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2017: 6450-6458.
[18]	HOU Y , MA Z , LIU C ,et al. Learning lightweight lane detection CNNs by self-attention distillation[C]// 2019 IEEE International Conference on Computer Vision. Piscataway:IEEE Press, 2019: 1013-1021.
[19]	ZHANG X , WANG T , QI J ,et al. Progressive attention guided recurrent network for salient object detection[C]// 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2018: 714-722.
[20]	LONG J , SHELHAMER E , DARRELL T . Fully convolutional networks for semantic segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017,39(4): 640-651.
[21]	HE K , GKIOXARI G , DOLLáR P ,et al. Mask R-CNN[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020,42(2): 386-397.
[22]	SZEGEDY C , LIU W , JIA Y ,et al. Going deeper with convolutions[C]// 2015 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2015: 1-9.
[23]	CHEN L C , PAPANDREOU G , SCHROFF F ,et al. Rethinking atrous convolution for semantic image segmentation[J]. arXiv Preprint,arXiv:1706.05587, 2017.
[24]	UIJLINGS J R R , VAN D S K E A , GEVERS T ,et al. Selective search for object recognition[J]. International Journal of Computer Vision, 2013,104(2): 154-171.
[25]	ZHU Z , LIANG D , ZHANG S ,et al. Traffic-sign detection and classification in the wild[C]// 2016 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2016: 2110-2118.
[26]	LARSSON F , FELSBERG M , FORSSEN P E . Correlating Fourier descriptors of local patches for road sign recognition[J]. IET Computer Vision, 2011,5(4): 244-254.
[27]	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.
[28]	SERMANET P , EIGEN D , ZHANG X ,et al. Overfeat:integrated recognition,localization and detection using convolutional networks[J]. arXiv Preprint,arXiv:1312.6229, 2013.

方法	CA	MGSA	Precision	Recall
YOLOv3	—	—	86.6%	89.4%
YOLOv3-CA	√	—	87.2%	90.3%
YOLOv3-MGSA	—	√	88.0%	91.6%
YOLOv3-A	√	√	88.5%	92.2%

YOLOv3-A：基于注意力机制的交通标志检测网络

YOLOv3-A: a traffic sign detection network based on attention mechanism

在线阅读

PDF下载

可视化

摘要/Abstract

引用本文

使用本文

图/表 18

参考文献 28

相关文章 11

Metrics

推荐阅读 0

[1]	冯海林, 张潇, 刘同存. 融合评论文本特征和评分图卷积表示的推荐模型[J]. 通信学报, 2022, 43(3): 164-171.
[2]	王泽宇, 布树辉, 黄伟, 郑远攀, 吴庆岗, 常化文, 张旭. 基于多模态特征的无监督领域自适应多级对抗语义分割网络[J]. 通信学报, 2022, 43(12): 157-171.
[3]	仲美玉, 吴培良, 窦燕, 刘毅, 孔令富. 基于中文语义-音韵信息的语音识别文本校对模型[J]. 通信学报, 2022, 43(11): 65-79.
[4]	陈卓, 朱淼, 杜军威. 基于多视角图神经网络的欺诈检测算法[J]. 通信学报, 2022, 43(11): 225-232.
[5]	王洪雁, 袁海. 基于骨骼及表观特征融合的动作识别方法[J]. 通信学报, 2022, 43(1): 138-148.
[6]	赵晓娟, 贾焰, 李爱平, 陈恺. 基于层级注意力机制的链接预测模型研究[J]. 通信学报, 2021, 42(3): 36-44.
[7]	杨军,党吉圣. 基于上下文注意力CNN的三维点云语义分割[J]. 通信学报, 2020, 41(7): 195-203.
[8]	李琳辉,周彬,连静,周雅夫. 基于社会注意力机制的行人轨迹预测方法研究[J]. 通信学报, 2020, 41(6): 175-183.
[9]	王竹荣, 薛伟, 牛亚邦, 崔颖安, 孙钦东, 黑新宏. 基于注意力机制的泊位占有率预测模型研究[J]. 通信学报, 2020, 41(12): 182-192.
[10]	李琳辉,钱波,连静,郑伟娜,周雅夫. 基于卷积神经网络的交通场景语义分割方法研究[J]. 通信学报, 2018, 39(4): 123-130.
[11]	黄凤辰,李敏,石爱业,汤敏,徐立中. 受昆虫视觉启发的多光谱遥感影像小目标检测[J]. 通信学报, 2011, 32(9): 88-95.