ZHONG Huijuan,CAI Qingyong.Traffic sign recognition method based on multi - scale convolutional neural network[J].Journal of Yanbian University,2020,46(04):359-365.
基于多尺度卷积神经网络的交通标志识别方法
- Title:
- Traffic sign recognition method based on multi - scale convolutional neural network
- 文章编号:
- 1004-4353(2020)04-0359-07
- 关键词:
- 交通标志识别; 卷积神经网络; TSR -MSCNN; 多尺度特征
- Keywords:
- traffic sign recognition; convolutional neural network; TSR -MSCNN; multi - scale features
- 分类号:
- TP391
- 文献标志码:
- A
- 摘要:
- 为了提升交通标志自动识别的精度,提出一种基于多尺度CNN的交通标志识别方法(TSR -MSCNN算法).该方法采用三阶段卷积神经网络,融合了低阶、中阶和高阶3种不同尺度的特征,并串联了多个小卷积层用以代替单个较大卷积层.通过对全连接层的神经元个数、Dropout参数、卷积核尺寸等网络超参数进行选比实验,获得了最佳的网络超参数.利用德国交通标志基准数据库(GTSRB)对不同算法进行测试表明,本文提出的算法在较小的网络参数量下能够有效提取交通标志特征,获取的识别准确率达到99.76%,且显著优于传统卷积神经网络方法和多尺度特征方法的识别准确率,因此本文算法在图像识别领域有良好的应用价值.
- Abstract:
- In order to improve the accuracy of automatic traffic sign recognition, we propose a traffic sign recognition algorithm based on multi -scale CNN. This method uses a three -stage convolutional neural network to fuse features of three different scales: low -order, medium -order, and high -order; and concatenates multiple small convolutional layers to replace a single larger convolution layer. In addition, the network hyperparameters such as the number of neurons in the fully connected layer, the dropout parameters, and the sizes of the convolution kernels are investigated and compared to obtain the best hyperparameter set. Different algorithms are tested on the German traffic sign recogninion benchmark(GTSRB). Experimental results show that the proposed algorithm in this paper can effectively extract traffic sign features and obtain recognition accuracy of 99.76% under a small amount of network parameters, which is obviously superior to the traditional convolutional neural network method and multi -scale feature recognition method. Therefore, the algorithm proposed in this paper has good usability in the field of image recognition.
参考文献/References:
[1] 周飞燕,金林鹏,董军.卷积神经网络研究综述[J].计算机学报,2017,40(6):1229-1251.
[2] 李新,禹翼.基于SIFT算法的交通标志识别[J].制造业自动化,2012,34(5):10-12.
[3] DOUVILLE P. Real -time classification of traffic signs[J]. Real -time Imaging, 2000,6(3):185-193.
[4] MALDONADO-BASCON S, LAFUENTE-ARROYO S, GIL -JIMENEZ P, et al. Road -sign detection and recognition based on support vector machines[J]. Intelligent Transportation Systems, 2007,8(2):264-278.
[5] 甘露,田丽华,李晨.基于融合特征和BP网络的交通标志识别方法[J].计算机工程与设计,2017,(38)10:2783-2813.
[6] SERMANET P, LECUN Y. Trafficsign recognition with multi -scale convolutional networks[C]//The 2011 International Joint Conference on Neural Networks(IJCNN). Washington DC: IEEE Computer Society, 2011:2809-2813.
[7] STALLKAMP J, SCHLIPSING M, SALMEN J, et al. The German traffic sign recognition benchmark[EB/OL].[2012-02-06].http://benchmark.ini.rub.de/?section=gtsrb&subsection=news.
[8] 王晓斌,黄金杰,刘文举.基于优化卷积神经网络结构的交通标志识别[J].计算机应用,2017,37(2):530-534.
[9] 宋青松,张超,田正鑫,等.基于多尺度卷积神经网络的交通标志识别[J].湖南大学学报(自然科学版),2018,45(8):131-137.
[10] SIMONYAN K, ZISSERMAN A. Very deep convolutional networks for large -scale image recognition[EB/OL].[2014-04-10].https://arxiv.org/abs/1409.1556.
[11] 仲会娟.基于CNN的多尺度特征在手写数字识别中的应用[J].绵阳师范学院学报,2019,11(5):22-26.
[12] LONG J, SHELHAMER E, DARRELL T. Fully convolutional networks for semantic segmentation[C]//IEEE Conference on Computer Vision and Pattern Recognition. Washington DC: IEEE Computer Society, 2015:3431-3440.
[13] 冀晓兵.图像尺寸变换的算法研究[D].西安:建筑科技大学信息与技术控制学院,2017.
[14] 陈清江,李毅,柴昱洲.一种基于深度学习的多聚焦图像融合算法[J].激光与光电子学进展,2018,55(7):246-254.
[15] ZEILER M, FERGUS R. Visualizing and under standing convolutional networks[C]//13th European conference on Computer Vision. Zurich: Springer, 2014:818-833.
备注/Memo
收稿日期: 2020-07-26 作者简介: 仲会娟(1985—),女,讲师,研究方向为图像与信号处理、无线通信技术.
基金项目: 福建省中青年教师教育科研项目(JT180724); 电子信息与通信技术慕课应用型团队项目(2019sjtd01)