基于改进LeNet-5网络的交通标志识别方法
|
摘要
针对传统LeNet-5卷积神经网络用于交通标志等多种类识别任务中,存在识别正确率低、网络容易过拟合以及梯度消失等问题进行改进。引入Inception卷积模块组来提取目标丰富的特征,同时增加网络的深度。引入BN (batch normalization)层对输入批量样本进行规范化处理;同时改用性能更好的Relu激活函数,并使用全局池化层代替全连接层,合理改变卷积核的大小和数目。研究结果表明,改进LeNet-5网络能够有效解决过拟合和梯度消失等问题,具有较好的鲁棒性;网络识别率达到98. 5%以上,相比CNN (convolutional neural network)+SVM (support vector machine)提高了约5%,比传统的LeNet-5网络提高了3%。可见,改进后的LeNet-5网络图像识别的准确率得到显著提高。
For the traditional LeNet-5 convolution neural network used in traffic signs and other kinds of recognition tasks,the problems such as low recognition accuracy,easy network over-fitting and gradient disappearance are improved. Inception convolution module group was cited to extract rich features of the target while increasing the depth of the network,the BN( batch normalization) layer was introduced to normalize the input batch samples to improve the input of the neural network. At the same time,the better Relu activation function was used,and the global pooling layer was used instead of the full connection layer,and the size and number of convolution kernels were reasonably changed. The research results show that the improved LeNet-5 network can effectively solve the problems of over-fitting and gradient disappearance,and has better robustness. At the same time,compared with CNN( convolutional neural network) + SVM( support vector machine) and traditional LeNet-5 network,the accuracy of the improved network classification can be up to 98. 5%,which is 5% higher than CNN + SVM and 3%higher than traditional LeNet-5 network. The accuracy of image recognition is improved significantly.
For the traditional LeNet-5 convolution neural network used in traffic signs and other kinds of recognition tasks,the problems such as low recognition accuracy,easy network over-fitting and gradient disappearance are improved. Inception convolution module group was cited to extract rich features of the target while increasing the depth of the network,the BN( batch normalization) layer was introduced to normalize the input batch samples to improve the input of the neural network. At the same time,the better Relu activation function was used,and the global pooling layer was used instead of the full connection layer,and the size and number of convolution kernels were reasonably changed. The research results show that the improved LeNet-5 network can effectively solve the problems of over-fitting and gradient disappearance,and has better robustness. At the same time,compared with CNN( convolutional neural network) + SVM( support vector machine) and traditional LeNet-5 network,the accuracy of the improved network classification can be up to 98. 5%,which is 5% higher than CNN + SVM and 3%higher than traditional LeNet-5 network. The accuracy of image recognition is improved significantly.
引文
1党倩,马苗,陈昱莅.基于二级改进的Le Net-5交通标志识别算法.陕西师范大学学报,2017;45(2):24-28Dang Qian,Ma Miao,Cheng Yuli.Traffic sign recognition algorithm based on two-stage improved Le Net-5.Journal of Shaanxi Normal University,2017;45(2):24-28
2郝岩,白艳萍,张校非,等.卷积神经网络在SAR目标识别中的应用.重庆理工大学学报,2018;32(5):204-209Hao Yan,Bai Yanping,Zhang Xiaofei,et al.Application of convolution neural network in SAR target recognition.Journal of Chongqing University of Technology,2018;32(5):204-209
3王洋.一种基于模板匹配的交通标志识别方法.长春:吉林大学,2013Wang Yang.A traffic sign recognition method based on template matching.Changchun:Jilin University,2013
4姜嘉琳.基于结构稀疏的目标识别方法.西安:电子科技大学,2013Jiang Jialin.Target recognition method based on sparse structure.Xi'an:University of Electronic Science and Technology,2013
5李新叶,黄腾.基于多尺度跃层卷积神经网络的精细车型识别.科学技术与工程,2017;17(11):246-249Li Xinye,Huang Teng.Fine vehicle identification based on multiscale convolution neural network.Science Technology and Engineering,2017;17(11):246-249
6胡杰,李少波,于丽娅,等.基于卷积神经网络与随机森林算法的专利文本分类模型.科学技术与工程,2018;18(6):268-272Hu Jie,Li Shaobo,Yu Liya,et al.Patent text classification model based on convolutional neural network and random forest algorithm.Science Technology and Engineering,2018;18(6):268-272
7李思琴,林磊,孙承杰.基于卷积神经网络的搜索广告点击率预测.智能计算机与应用,2015;5(5):22-28Li Siqin,Lin Lei,Sun Chengjie.Click rate prediction of search ads based on convolution neural network.Intelligent Computers and Applications,2015;5(5):22-28
8吴阳阳,彭广德,吴相飞.基于Le Net-5改进的卷积神经网络图像识别方法.信息与电脑,2018;(7):127-130Wu Yangyang,Peng Guangdei,Wu Xiangfei.Image recognition method based on improved Le Net-5 convolution neural network.Information and Computer,2018;(7):127-130
9王秀席,王茂宁,张建伟,等.基于改进的卷积神经网络Le Net-5的车型识别方法.计算机应用研究,2018;35(7):2215-2218Wang Xiuxi,Wang Maoning,Zhang Jianwei,et al.Vehicle identification method based on improved convolutional neural network Le Net-5.Computer Application Research,2018;35(7):2215-2218
10李丹,沈夏炯,张海香.基于Lenet-5的卷积神经网络改进算法.计算机时代,2016;30(8):4-6,12Li Dan,Shen Xiajiong,Zhang Haixiang.Improved convolution neural network algorithm based on lenet-5.Computer Generation,2016;30(8):4-6,12
2郝岩,白艳萍,张校非,等.卷积神经网络在SAR目标识别中的应用.重庆理工大学学报,2018;32(5):204-209Hao Yan,Bai Yanping,Zhang Xiaofei,et al.Application of convolution neural network in SAR target recognition.Journal of Chongqing University of Technology,2018;32(5):204-209
3王洋.一种基于模板匹配的交通标志识别方法.长春:吉林大学,2013Wang Yang.A traffic sign recognition method based on template matching.Changchun:Jilin University,2013
4姜嘉琳.基于结构稀疏的目标识别方法.西安:电子科技大学,2013Jiang Jialin.Target recognition method based on sparse structure.Xi'an:University of Electronic Science and Technology,2013
5李新叶,黄腾.基于多尺度跃层卷积神经网络的精细车型识别.科学技术与工程,2017;17(11):246-249Li Xinye,Huang Teng.Fine vehicle identification based on multiscale convolution neural network.Science Technology and Engineering,2017;17(11):246-249
6胡杰,李少波,于丽娅,等.基于卷积神经网络与随机森林算法的专利文本分类模型.科学技术与工程,2018;18(6):268-272Hu Jie,Li Shaobo,Yu Liya,et al.Patent text classification model based on convolutional neural network and random forest algorithm.Science Technology and Engineering,2018;18(6):268-272
7李思琴,林磊,孙承杰.基于卷积神经网络的搜索广告点击率预测.智能计算机与应用,2015;5(5):22-28Li Siqin,Lin Lei,Sun Chengjie.Click rate prediction of search ads based on convolution neural network.Intelligent Computers and Applications,2015;5(5):22-28
8吴阳阳,彭广德,吴相飞.基于Le Net-5改进的卷积神经网络图像识别方法.信息与电脑,2018;(7):127-130Wu Yangyang,Peng Guangdei,Wu Xiangfei.Image recognition method based on improved Le Net-5 convolution neural network.Information and Computer,2018;(7):127-130
9王秀席,王茂宁,张建伟,等.基于改进的卷积神经网络Le Net-5的车型识别方法.计算机应用研究,2018;35(7):2215-2218Wang Xiuxi,Wang Maoning,Zhang Jianwei,et al.Vehicle identification method based on improved convolutional neural network Le Net-5.Computer Application Research,2018;35(7):2215-2218
10李丹,沈夏炯,张海香.基于Lenet-5的卷积神经网络改进算法.计算机时代,2016;30(8):4-6,12Li Dan,Shen Xiajiong,Zhang Haixiang.Improved convolution neural network algorithm based on lenet-5.Computer Generation,2016;30(8):4-6,12