基于双通路跃层卷积网络的交通标志识别算法
|
摘要
交通标志识别(Traffic Sign Recognition,TSR)是智能交通系统的重要研究方向之一。因道路交通的环境复杂、交通标志数据库规模大小等因素制约,在设计TSR系统可行性方案时必须考虑算法的复杂度、识别率和鲁棒性。针对这一问题,本文提出了一种不同尺度的双通路跃层卷积神经网络算法,在同一通路上交通标志的底层局部特征和高层全局的特征,与不同通路上经过局部响应归一化和池化后的特征在全连接层融合,从而丰富了交通标志分类的特征,最后将特征图输入分类器进行交通标志识别。采用德国交通标志识别标准数据集(German Traffic Sign Recognition Benchmark,GTSRB)进行训练和测试,本文算法的识别率达到97.96%,明显优于单一通路的跃层卷积网络算法和人工方法。
Traffic Sign Recognition(TSR) is one of the important directions in Intelligent Transportation System research.Restricted by the complex traffic environment, the traffic sign database scale and other factors, it is necessary to consider the complexity of the algorithm, recognition rate, and robustness. Concerning this issue, we propose Double Channels Layer-Skipping Convolutional Neural Network(DCLS-CNN) structure at different scales. The low-layer local features and high-layer global features of traffic signs on one channel, and other channel feature by the Local Response Normalization and pooling, which enriched the feature of traffic signs classification, these features fusion in the full connected layer,finally features were exported for traffic signs recognition. Tested by the German Traffic Sign Recognition Benchmark(GTSRB), the recognition rate of this algorithm is up to 97.96%, which is significantly higher than that of Layer-Skipping Convolutional Neural Network(LS-CNN) and hand-crafted methods.
Traffic Sign Recognition(TSR) is one of the important directions in Intelligent Transportation System research.Restricted by the complex traffic environment, the traffic sign database scale and other factors, it is necessary to consider the complexity of the algorithm, recognition rate, and robustness. Concerning this issue, we propose Double Channels Layer-Skipping Convolutional Neural Network(DCLS-CNN) structure at different scales. The low-layer local features and high-layer global features of traffic signs on one channel, and other channel feature by the Local Response Normalization and pooling, which enriched the feature of traffic signs classification, these features fusion in the full connected layer,finally features were exported for traffic signs recognition. Tested by the German Traffic Sign Recognition Benchmark(GTSRB), the recognition rate of this algorithm is up to 97.96%, which is significantly higher than that of Layer-Skipping Convolutional Neural Network(LS-CNN) and hand-crafted methods.
引文
[1]Hamed H.A,Elnaz J.H,Domenec P.A practical approach for detection and classification of traffic signs using Convolutional Neural Networks[J].Robotics and AutonOmous Systems 2016(84):97-112.
[2]Greenhalgh J,Mirmehdi M.Real-time detection and recognition of road traffic signs[J].IEEE Transactions on Intelligent Transportation Systems,2012,13(4):1498-1506.
[3]Liu C,Yuen J,Torralba A.Sift flow:Dense correspondence across scenes and its applications[J].IEEE transactions on pattern analysis and machine intelligence,2011,33(5):978-994.
[4]Sánchez J,Perronnin F,Mensink T,et al.Image classification with the fisher vector:Theory and practice[J].International journal of computer vision,2013,105(3):222-245.
[5]Shalev-Shwartz S,Singer Y,Srebro N,et al.Pegasos:Primal estimated sub-gradient solver for svm[J].Mathematical programming,2011,127(1):3-30.
[6]A.Ellahyani,M.E.Ansari,I.E.Jaafari.Traffic sign detection and recognition based on random forests[J].Applied soft Computing 2016,46(9):805-815.
[7]Y.Tan,P.Tang,Y.Zhou,W.Luo,Y.Kang,G.Li.Photograph aesthetical evaluation and classification with deep convolutional neural networks[J].Neurocomputing 2017,228(3):165-175.
[8]Redmon J,FarhadiA.YOL09000:Better,faster,stronger[EB/OL].[2070-09-20].https://arxiv.org/abs/1612.08242.
[9]He K,Zhang X,Ren S,et al.Delving deep into rectifiers:Surpassing human-level performance on imagenet classification[C].Proceedings of the IEEE international conference on computer vision.2015:1026-1034.
[10]Scherer D,Müller A,Behnke S.Evaluation of pooling operations in convolutional architectures for object recognition[C].International conference on artificial neural networks.2010:92-101.
[11]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.
[12]谭帧刚.基于多分辨率卷积网络的交通标志检测与识别[D].大连:大连理工大学,2015.
[13]Cires,An D,Meier U,Masci J,et al.Multi-column deep neural network for traffic sign classification[J].Neural Networks,2012(32):333-338.
[14]Zaklouta F,Stanciulescu B.Real-time traffic-sign recognition using tree classifiers[J].IEEE Transactions on Intelligent Transportation Systems,2012,13(4):1507-1514.
[15]Gonzalez-Reyna S E,Avina-Cervantes J G,Ledesma-Orozco S E,et al.Traffic Sign Recognition Based on Linear Discriminant Analysis[C].MICAI2013(2):185-193.
[16]Sermanet P,LeCun Y.Traffic sign recognition with multi-scale convolutional networks[C].Neural Networks(IJCNN),The 2011 International Joint Conference on.IEEE,2011:2809-2813.
[2]Greenhalgh J,Mirmehdi M.Real-time detection and recognition of road traffic signs[J].IEEE Transactions on Intelligent Transportation Systems,2012,13(4):1498-1506.
[3]Liu C,Yuen J,Torralba A.Sift flow:Dense correspondence across scenes and its applications[J].IEEE transactions on pattern analysis and machine intelligence,2011,33(5):978-994.
[4]Sánchez J,Perronnin F,Mensink T,et al.Image classification with the fisher vector:Theory and practice[J].International journal of computer vision,2013,105(3):222-245.
[5]Shalev-Shwartz S,Singer Y,Srebro N,et al.Pegasos:Primal estimated sub-gradient solver for svm[J].Mathematical programming,2011,127(1):3-30.
[6]A.Ellahyani,M.E.Ansari,I.E.Jaafari.Traffic sign detection and recognition based on random forests[J].Applied soft Computing 2016,46(9):805-815.
[7]Y.Tan,P.Tang,Y.Zhou,W.Luo,Y.Kang,G.Li.Photograph aesthetical evaluation and classification with deep convolutional neural networks[J].Neurocomputing 2017,228(3):165-175.
[8]Redmon J,FarhadiA.YOL09000:Better,faster,stronger[EB/OL].[2070-09-20].https://arxiv.org/abs/1612.08242.
[9]He K,Zhang X,Ren S,et al.Delving deep into rectifiers:Surpassing human-level performance on imagenet classification[C].Proceedings of the IEEE international conference on computer vision.2015:1026-1034.
[10]Scherer D,Müller A,Behnke S.Evaluation of pooling operations in convolutional architectures for object recognition[C].International conference on artificial neural networks.2010:92-101.
[11]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.
[12]谭帧刚.基于多分辨率卷积网络的交通标志检测与识别[D].大连:大连理工大学,2015.
[13]Cires,An D,Meier U,Masci J,et al.Multi-column deep neural network for traffic sign classification[J].Neural Networks,2012(32):333-338.
[14]Zaklouta F,Stanciulescu B.Real-time traffic-sign recognition using tree classifiers[J].IEEE Transactions on Intelligent Transportation Systems,2012,13(4):1507-1514.
[15]Gonzalez-Reyna S E,Avina-Cervantes J G,Ledesma-Orozco S E,et al.Traffic Sign Recognition Based on Linear Discriminant Analysis[C].MICAI2013(2):185-193.
[16]Sermanet P,LeCun Y.Traffic sign recognition with multi-scale convolutional networks[C].Neural Networks(IJCNN),The 2011 International Joint Conference on.IEEE,2011:2809-2813.