基于伪样本正则化Faster R⁃CNN的交通标志检测

doi:10.13229/j.cnki.jdxbgxb20200956

吉林大学学报(工学版) ›› 2021, Vol. 51 ›› Issue (4): 1251-1260.doi: 10.13229/j.cnki.jdxbgxb20200956

• 交通运输工程·土木工程 • 上一篇

基于伪样本正则化Faster R⁃CNN的交通标志检测

李厚杰¹(),王法胜¹(),贺建军¹,周瑜¹,李威²,窦宇轩¹

^1.大连民族大学信息与通信工程学院，辽宁大连 116605
^2.大连民族大学计算机科学与工程学院，辽宁大连 116605

收稿日期:2020-12-10 出版日期:2021-07-01 发布日期:2021-07-14
通讯作者: 王法胜 E-mail:lhj@dlnu.edu.cn;wangfasheng@dlnu.edu.cn
作者简介:李厚杰（1977-），男，副教授，博士. 研究方向：计算机视觉，交通图像处理. E-mail： lhj@dlnu.edu.cn
基金资助:
国家自然科学基金项目(61972068);辽宁省自然科学基金项目(20180550625);辽宁省“百千万人才工程”项目(2018B09);大连市青年科技之星项目(2017RQ151);辽宁省高校创新人才计划项目(LR2019020);兴辽英才计划项目(XLYC2007023)

Pseudo sample regularization Faster R⁃CNN for traffic sign detection

Hou⁃jie LI¹(),Fa⁃sheng WANG¹(),Jian⁃jun HE¹,Yu ZHOU¹,Wei LI²,Yu⁃xuan DOU¹

^1.School of Information and Communication Engineering，Dalian Minzu University，Dalian 116605，China
^2.School of Computer Science and Engineering，Dalian Minzu University，Dalian 116605，China

Received:2020-12-10 Online:2021-07-01 Published:2021-07-14
Contact: Fa?sheng WANG E-mail:lhj@dlnu.edu.cn;wangfasheng@dlnu.edu.cn

摘要/Abstract

摘要：

针对交通标志检测算法往往仅能对特定类标志检测或基于深度学习方法因训练样本少而造成“过拟合”高风险等问题，本文提出了一种基于伪样本正则化Faster R?CNN深度学习的标志检测算法。该算法首先通过训练数据集提供的标志和无标注的背景样本，提出了一种伪样本正则化策略。然后，通过深度学习模型中区域建议生成网络获取建议区域。最后，利用交替训练策略、共享CNN策略和联合训练策略、RPN网络和Fast R?CNN目标检测网络交替训练和联合训练，最终获取Faster R?CNN交通标志检测模型，实现了各类标志的检测，并有效降低了“过拟合”风险。实验结果验证了本文算法的有效性。

关键词: 信息处理技术, 交通标志检测, 伪样本正则化, Faster R?CNN, 区域建议网络

Abstract:

Traffic sign detection is the key step of a traffic sign recognition system. To solve the problems of only detection specific categories of traffic signs and over?fitting due to the lack of training samples in deep learning based methods， we propose a deep learning based traffic sign detection method ? pseudo sample regularization Faster R?CNN. In our method， we first design a pseudo sample regularization scheme using the traffic signs and unlabeled background samples from the training set. Then， the region proposal network （RPN） in the deep learning framework is used to generate region proposals. At last， the RPN network and Fast R?CNN detection network are working alternatively and jointly using the alternative training， shared CNN and co?training strategies to obtain the Faster R?CNN traffic sign detection model. We demonstrate the effectiveness of the proposed method through comprehensive experimental analysis. The results show that our method shows boosted traffic sign detection performance and the over?fitting problem can be suppressed。

Key words: information processing technology, traffic sign detection, pseudo sample regularization, faster region?convolutional neural networks （Faster R?CNN）, region proposal networks

中图分类号:

TN911.73

李厚杰,王法胜,贺建军,周瑜,李威,窦宇轩. 基于伪样本正则化Faster R⁃CNN的交通标志检测[J]. 吉林大学学报(工学版), 2021, 51(4): 1251-1260.

Hou⁃jie LI,Fa⁃sheng WANG,Jian⁃jun HE,Yu ZHOU,Wei LI,Yu⁃xuan DOU. Pseudo sample regularization Faster R⁃CNN for traffic sign detection[J]. Journal of Jilin University(Engineering and Technology Edition), 2021, 51(4): 1251-1260.

图/表 8

图1

表1

图2

表2

不同正则化条件下的检测性能比较（AP） (%)"

禁止类	警告类	指示类	其他类	$m A P$
90.85	99.59	99.247	100	97.42
99.89	90.91	98.64	100	97.35
99.83	99.09	99.81	100	99.68
100	99.72	99.45	100	99.79

表2

图3

表3

各种检测方法在GTSDB上的检测性能比较（AP） (%)"

方法	禁止类	警告类	指示类	其他类	$m A P$
［26］	100	99.91	100.00	?	?
［27］	100	98.85	92.00	?	?
［13］	99.98	98.85	95.76	?	?
［28］	99.29	97.13	96.74	?	?
［29］	?	99.73	97.62	?	?
［23］	96.81	96.12	94.02	?	?
［24］	99.89	99.90	99.16	?	?
本文方法	99.83	99.09	99.82	100%	99.68%

表3

图4

图5

参考文献 29

1	Tabernik D, Skoaj D. Deep learning for large⁃scale traffic⁃sign detection and recognition[J]. IEEE Transactions on Intelligent Transportation Systems, 2019, 21(4): 1427⁃1440.
2	Li H J, Qiu T S, Song H Y, et al. A fast traffic signs detection method based on color segmentation and improved radial.symmetry[J]. ICIC Express Letters, 2014, 8(8): 2175⁃2180.
3	Temel D, Chen M H, Alregib G. Traffic sign detection under challenging conditions: a deeper look into performance variations and spectral characteristics[J]. IEEE Transactions on Intelligent Transportation Systems, 2020, 21: 3663⁃3673.
4	王方石, 王坚, 李兵,等. 基于深度属性学习的交通标志检测[J]. 吉林大学学报:工学版, 2018,48(1):319-329.
	Fang-shi WANG, Jian WANG, Bing LI, et al. Deep attribute learning based traffic sign detection[J].Journal of Jilin University(Engineering and Technology Edition)2018, 48(1):319-329.
5	Gudigar A, Chokkadi S, Raghavendra U. A review on automatic detection and recognition of traffic sign[J]. Multimedia Tools &Applications, 2016, 75(1): 1⁃32.
6	Gonzalez A, Garrido M A, Llorca D F, et al. Automatic traffic signs and panels inspection system using computer vision[J]. IEEE Transactions on Intelligent Transportation Systems, 2011, 12(1): 485⁃499.
7	Youssef A, Albani D, Nardi D, et al. Fast traffic sign recognition using color segmentation and deep convolutional networks[C]∥International Conference on Advanced Concepts for Intelligent Vision Systems, Lecce, Puglia, Italy, 2016: 205⁃216.
8	Lafuente⁃arroyo S, Salcedo⁃sanz S, Maldonado⁃basc, et al. A decision support system for the automatic management of keep⁃clear signs based on support vector machines and geo⁃graphic information systems[J]. Expert Systems with Applications, 2010, 37(1): 767⁃773.
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	Cao J, Zhang J, Huang W. Traffic sign detection and recognition using multi⁃scale fusion and prime sample attention[J]. IEEE Access, 2020, 9: 3579⁃3591.
11	Yali B, Saadna Y. A fast and robust traffic sign recognition[J]. Issr Journals, 2014, 5(1): 139⁃149.
12	李厚杰, 邱天爽, 宋海玉, 等. 基于径向对称变换的自适应交通禁止标志检测[J]. 光电子·激光, 2014, 25(3):532⁃539.
	Li Hou⁃jie, Qiu Tian⁃shuang, Song Hai⁃yu, et al. Adaptive traffic prohibitive sign detection based on radial symmetry transform[J]. Journal of Optoelectronics Laser, 2014, 25(3): 532⁃539.
13	Salti S, Petrelli A, Tombari F, et al. A traffic sign detection pipeline based on interest region extraction[C]∥IEEE International Joint Conference on Neural Networks, Dallas, TX, USA, 2013: 1⁃7.
14	Sugiharto A, Harjoko A. Traffic sign detection based on HOG and PHOG using binary SVM and k⁃NN[C]∥IEEE International Conference on Information Technology, Computer, and Electrical Engineering, Semarang, Indonesia, 2017: 317⁃321.
15	Shi J H, Lin H Y. A vision system for traffic sign detection and recognition[C]∥IEEE 26th International Symposium on Industrial Electronics, Edinburgh, UK. 2017: 1596⁃1601.
16	Nguyen K D, Le D D, Duong D A. Efficient traffic sign detection using bag of visual words and multi⁃scales SIFT[C]∥International Conference on Neural Information Processing, Daegu, Korea, 2013: 433⁃441.
17	Zang D, Zhang J, Zhang D, et al. Traffic sign detection based on cascaded convolutional neural networks[C]∥IEEE/ACIS International Conference on Software Engineering, Artificial Intelligence, Networking and Parallel/Distributed Computing, Shanghai, China, 2016: 201⁃206.
18	Ellahyani A, Ansari M E, Jaafari I E. Traffic sign detection and recognition based on random forests[J]. Applied Soft Computing, 2016, 46: 805⁃815.
19	Kaplan K, Kurtul C, Akin H L. Real⁃time traffic sign detection and classification method for intelligent vehicles[C]∥IEEE International Conference on Vehicular Electronics and Safety, Istanbul, Turkey, 2012: 448⁃453.
20	Zhang K, Sheng Y, Li J. Automatic detection of road traffic signs from natural scene images based on pixel vector and central projected shape feature[J]. IET Intelligent Transport Systems, 2012, 6(3): 282⁃291.
21	Xiong C, Wang C, Ma W, et al. A traffic sign detection algorithm based on deep convolutional neural network[C]∥IEEE International Conference on Signal and Image Processing, Beijing, China, 2017: 676⁃679.
22	Zuo Z, Yu K, Zhou Q, et al. Traffic signs detection based on Faster R⁃CNN[C]∥IEEE 37th International Conference on Distributed Computing Systems Workshops, Atlanta, GA, USA, 2017: 286⁃288.
23	Zhang J, Huang M, Jin X, et al. A real⁃time Chinese traffic sign detection algorithm based on modified YOLOv2[J]. Algorithm, 2017, 10(4): a10040127.
24	Lee H, Kim K. Simultaneous traffic sign detection and boundary estimation using convolutional neural network[J]. IEEE Transactions on Intelligent Transportation Systems, 2018, 19(5): 1652⁃1663.
25	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 Machine Intelligence, 2017, 39(6): 1137⁃1149.
26	Wang G, Ren G, Wu Z, et al. A robust, coarse⁃to⁃fine traffic sign detection method[C]∥IEEE International Joint Conference on Neural Networks, Dallas, TX, USA, 2013: 1⁃5.
27	Liang M, Yuan M, Hu X, et al. Traffic sign detection by ROI extraction and histogram features⁃based recognition[C]∥The International Joint Conference on Neural Networks, Dallas, TX, USA, 2013: 1⁃8.
28	Yang Y, Luo H, Xu H, et al. Towards real⁃time traffic sign detection and classification[J]. IEEE Transactions on Intelligent Transportation Systems, 2016, 17(71): 2022⁃2031.
29	Wu Y, Liu Y, Li J, et al. Traffic sign detection based on convolutional neural networks[C]∥IEEE International Joint Conference on Neural Networks, Dallas, TX, USA, 2014: 1⁃7.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed

[1]	蒋华伟,杨震,张鑫,董前林. 图像去雾算法研究进展[J]. 吉林大学学报(工学版), 2021, 51(4): 1169-1181.
[2]	金静,党建武,王阳萍,申东. 融合模糊统计纹理特征的多线索粒子滤波跟踪[J]. 吉林大学学报(工学版), 2021, 51(3): 1111-1120.
[3]	郭继昌,乔珊珊. 基于深度图的水下图像复原[J]. 吉林大学学报(工学版), 2021, 51(2): 677-684.
[4]	刘国华,周文斌. 基于卷积神经网络的脉搏波时频域特征混叠分类[J]. 吉林大学学报(工学版), 2020, 50(5): 1818-1825.
[5]	史再峰,李金卓,曹清洁,李慧龙,胡起星. 基于生成对抗网络的低剂量能谱层析成像去噪算法[J]. 吉林大学学报(工学版), 2020, 50(5): 1755-1764.
[6]	王柯俨,王迪,赵熹,陈静怡,李云松. 基于卷积神经网络的联合估计图像去雾算法[J]. 吉林大学学报(工学版), 2020, 50(5): 1771-1777.
[7]	谌华,郭伟,闫敬文,卓文浩,吴良斌. 基于深度学习的SAR图像道路识别新方法[J]. 吉林大学学报(工学版), 2020, 50(5): 1778-1787.
[8]	张薇,韩勇,金铭,乔晓林. 基于托普利兹矩阵集重构的相干信源波达方向估计[J]. 吉林大学学报(工学版), 2020, 50(2): 703-710.
[9]	程艳芬,姚丽娟,袁巧,陈先桥. 水下视频图像清晰化方法[J]. 吉林大学学报(工学版), 2020, 50(2): 668-677.
[10]	于晓辉,张志成,李新波,孙晓东. 基于状态空间模型的指数衰减正弦信号参数估计[J]. 吉林大学学报(工学版), 2019, 49(6): 2083-2088.
[11]	刘富, 权美静, 王柯, 刘云, 康冰, 韩志武, 侯涛. 仿蝎子振源定位机理的位置指纹室内定位方法[J]. 吉林大学学报(工学版), 2019, 49(6): 2076-2082.
[12]	马子骥,卢浩,董艳茹. 双通道单图像超分辨率卷积神经网络[J]. 吉林大学学报(工学版), 2019, 49(6): 2089-2097.
[13]	郭继昌,吴洁,郭春乐,朱明辉. 基于残差连接卷积神经网络的图像超分辨率重构[J]. 吉林大学学报(工学版), 2019, 49(5): 1726-1734.
[14]	曹运合,曾丽,王宇. 基于特征空间的子阵级自适应和差波束测角方法[J]. 吉林大学学报(工学版), 2019, 49(5): 1735-1744.
[15]	卢洋,王世刚,赵文婷,赵岩. 基于离散Shearlet类别可分性测度的人脸表情识别方法[J]. 吉林大学学报(工学版), 2019, 49(5): 1715-1725.

基于伪样本正则化Faster R⁃CNN的交通标志检测

Pseudo sample regularization Faster R⁃CNN for traffic sign detection

RICH HTML

PDF (PC)

摘要/Abstract

引用本文

使用本文

图/表 8

参考文献 29

相关文章 15

Metrics

本文评价

推荐阅读 0