您好,欢迎来到中国测试科技资讯平台!

首页> 《中国测试》期刊 >本期导读>应用双通道卷积神经网络的交通标识识别方法

应用双通道卷积神经网络的交通标识识别方法

296    2024-06-26

¥0.50

全文售价

作者:赵泽毅, 周福强, 王少红, 徐小力

作者单位:北京信息科技大学 现代测控技术教育部重点实验室, 北京 100192


关键词:交通标识识别;LeNet-5网络结构;卷积神经网络


摘要:

针对交通标识识别问题,传统的LeNet-5网络结构对于交通标识识别准确率低,识别速度慢,并且忽略天气等自然因素的影响。通过卷积神经网络技术,提出一种基于LeNet-5改进的双通道、多尺度的网络结构模型。在双通道结构中每个通道包含两个分支结构,且每个通道的卷积个数和图像尺度不同,通过不同尺度图像的特征提取,使图像特征变得更为丰富。其次,改进后的网络结构大大增加卷积核的个数。最后,通过更改Sigmoid激活函数为ReLu激活函数,更改随机梯度下降算法为Adam算法,并添加Dropout层来防止过拟合,从而提高交通标识识别率。改进网络的识别率为98.6 %,上下浮动0.5 %,相对与传统的LeNet-5网络结构,识别率提高15 %以上,验证得出改进的网络结构具有一定的鲁棒性。


Traffic sign recognition method based on dual channel CNN
ZHAO Zeyi, ZHOU Fuqiang, WANG Shaohong, XU Xiaoli
Key Laboratory of Modern Measurement & Control Technology (Ministry of Education), Beijing Information Science and Technology University, Beijing 100192, China
Abstract: Aiming at the problem of traffic sign recognition, the traditional LeNet-5 network structure has low accuracy and slow recognition, and ignores the influence of natural factors such as weather. Through convolution neural network technology, an improved two channel and multi-scale network structure model based on LeNet-5 is proposed. In the dual channel structure, each channel contains two branch structures, and the number of convolutions of each channel is different from the image scale. Through the feature extraction of images with different scales, the image features become richer. Secondly, the improved network structure greatly increases the number of convolution cores. Finally, change the Sigmoid activation function to Relu activation function, change the random gradient descent algorithm to Adam algorithm, and add dropout layer to prevent over fitting, so as to improve the traffic sign recognition rate. The recognition rate of the improved network is 98.6 %, floating up and down by 0.5 %. Compared with the traditional LeNet-5 network structure, the recognition rate is increased by more than 15 %. It is verified that the improved network structure has certain robustness.
Keywords: traffic sign recognition; LeNet-5 network structure; CNN
2024, 50(6):35-41,48 收稿日期: 2022-07-18;收到修改稿日期: 2022-12-02
基金项目:
作者简介: 赵泽毅(1994-),男,河南洛阳市人,硕士研究生,专业方向为计算机视觉。
参考文献
[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.
ZHOU F Y, JIN L P, DONG J. A review of convolutional neural network research [J].Chinese Journal of Computers, 2017, 40 (6): 1229-1251.
[2] 孙志军, 薛磊, 许阳明, 等. 深度学习研究综述[J]. 计算机应用研究, 2012, 29(8): 2806-2810.
SUN Z J, XUE L, XU Y M, et al.A review of deep learning research [J].Application Research of Computers 2012, 29 (8): 2806-2810.
[3] 段建, 翟慧敏. 深度卷积神经网络在Caltech-101图像分类中的相关研究[J]. 计算机应用与软件, 2016, 33(12): 165-168.
DUAN J, ZHAI H M. Research on deep convolutional neural networks in Caltech 101 image classification [J].Computer Applications and Software, 2016, 33 (12): 165-168.
[4] 王昕, 赵飞, 蒋佐富, 等. 迁移学习和卷积神经网络电力设备图像识别方法[J]. 中国测试, 2020, 46(5): 108-113.
WANG X, ZHAO F, JIANG Z F, et al. Power equipment image recognition method based on transfer learning and convolutional neural network[J]. China Measurement & Test, 2020, 46 (5): 108-113.
[5] 李英杰, 张惊雷. 基于全卷积网络的图像语义分割算法[J]. 计算机应用与软件, 2020, 37(8): 213-218.
LI Y J, ZHANG J L. Image semantic segmentation algorithm based on fully convolutional networks [J]. Computer Applications and Software, 2020, 37 (8): 213-218.
[6] 顾攀, 张烽栋. 基于神经网络的图像弱监督语义分割算法[J]. 计算机应用与软件, 2018, 35(2): 284-288.
GU P, ZHANG F D. A weakly supervised semantic segmentation algorithm for images based on neural networks [J].Computer Applications and Software, 2018, 35 (2): 284-288.
[7] 彭晏飞, 陶进, 訾玲玲. 基于卷积神经网络和E2LSH的遥感图像检索研究[J]. 计算机应用与软件, 2018, 35(7): 250-255.
PENG Y F, TAO J, ZI L L. Research on remote sensing image retrieval based on convolutional neural networks and E2LSH [J].Computer Applications and Software, 2018, 35 (7): 250-255.
[8] 彭伟康, 陈爱军, 吴东明, 等. 基于改进Faster R-CNN的水准泡缺陷检测方法[J]. 中国测试, 2021, 47(7): 6-12.
PENG W K, CHEN A J, WU D M, et al.Defect detection method of level bubble based on improved Faster R-CNN[J]. China Measurement & Test, 2021, 47 (7): 6-12.
[9] 郭雨青, 曾庆军, 夏楠, 等. 图像增强水下自主机器人目标识别研究[J]. 中国测试, 2021, 47(11): 47-52.
GUO Y Q, ZENG Q J, XIA N, et al. Research on target recognition of autonomous underwater vehicle based on image enhancement[J]. China Measurement & Test, 2021, 47 (11): 47-52.
[10] 揭伟, 李为相, 李为. 基于多特征融合的交通标识实时分类与识别[J]. 现代电子技术, 2019, 42(11): 50-53.
JIE W, LI W X, LI W .Real time classification and recognition of traffic signs based on multi feature fusion [J].Modern Electronic Technology, 2019, 42 (11): 50-53.
[11] 齐朗晔, 张重阳, 何成东. 基于多特征组合的交通标识识别[J]. 计算机工程与科学, 2015, 37(4): 776-782.
QI L Y, ZHANG C Y, HE C D. Traffic sign recognition based on multi feature combination [J] .Computer Engineering & Science, 2015, 37 (4): 776-782.
[12] 邓翔宇, 张屹南, 杨雅涵. 一种用于交通标识分类的形状识别算法[J]. 计算机工程与科学, 2021, 43(2): 322-328.
DENG X Y, ZHANG Y N, YANG Y H. A shape recognition algorithm for traffic sign classification [J]. Computer Engineering & Science, 2021, 43 (2): 322-328.
[13] 赵志宏, 杨绍普, 马增强. 基于卷积神经网络LeNet-5的车牌字符识别研究[J]. 系统仿真学报, 2010, 22(3): 638-641.
ZHAO Z H, YANG S P, MA Q Q. Research on license plate character recognition based on convolutional neural network LeNet-5 [J].Journal of System Simulation, 2010, 22 (3): 638-641.
[14] 王秀席, 王茂宁, 张建伟, 等. 基于改进的卷积神经网络LeNet-5的车型识别方法[J]. 计算机应用研究, 2018, 35(7): 2215-2218.
WANG X X, WANG M N, ZHANG J W, et al. A vehicle recognition method based on improved convolutional neural network LeNet-5 [J].Application Research of Computers, 2018, 5 (7): 2215-2218.
[15] SERMANET P, LECUN Y. Traffic sign recognition with multi-scale Convolutional Networks[C]// International Joint Conference on Neural Networks. IEEE, 2011: 2809-2813.
[16] 张焕, 张庆, 于纪言. 激活函数的发展综述及其性质分析[J/OL]. 西华大学学报(自然科学版): 1-10.1-08-11.
ZHANG H, ZHANG Q, YU J Y. Development review and property analysis of activation functions [J/OL]. Journal of Xihua University (Natural Science Edition): 1-10.1-08-11.
[17] 陈超, 齐峰. 卷积神经网络的发展及其在计算机视觉 领域中的应用综述[J]. 计算机科学, 2019, 46(3): 63-73.
CHEN C, QI F. A review of the development of convolutional neural networks and their applications in the field of computer vision [J].Computer Science, 2019, 46 (3): 63-73.
[18] 杨观赐, 杨静, 李少波, 等. 基于Dropout与ADAM优化器的改进CNN算法[J]. 华中科技大学学报(自然科学版), 2018, 46(7): 122-127.
YANG G C, YANG J, LI S B, et al.An improved CNN algorithm based on Dropout and ADAM optimizer [J].Journal of Huazhong University of Science and Technology(Natural Science Edition), 2018, 46 (7): 122-127.
[19] 王功鹏, 段萌, 牛常勇. 基于卷积神经网络的随机梯度下降算法[J]. 计算机工程与设计, 2018, 39(2): 441-445.
WANG G P, DUAN M, NIU C Y. A random gradient descent algorithm based on convolutional neural networks [J].Computer Engineering and Design, 2018, 39 (2): 441-445.