使用多任务级联卷积神经网络进行车牌照识别
|
摘要
车牌照识别是计算机视觉技术的一项重要应用,在保障日常交通秩序中扮演关键角色,因此,一直以来车牌照识别技术都是计算机视觉领域中一项热门研究。由于车牌照识别系统的应用场景通常在交通路口,交通情况复杂度高,由于车辆运动和拍摄角度导致的遮挡、图像变形等问题,使得高精度、抗干扰的车牌照识别系统具有重要研究价值。近年来,随着深度学习技术的发展,特别是其在图像识别和目标检测领域带来相较传统方法的巨大的性能提升。文章尝试使用深度卷积神经网络的方法,提出了一个在精确度和鲁棒性方面都较为出色的针对中国大陆车牌照的识别算法,其在我们的私有数据集上识别准确率达到94.93%,识别图片平均耗时243 ms。该方法可以通过更换数据集来泛化到其他国家或地区的车牌照识别问题上。
Vehicle license plate recognition is an important application of computer vision technology and plays a key role in ensuring daily traffic order.Therefore,license plate recognition technology has always been a hot research in the field of computer vision.Since the application scene of the license plate recognition system is usually at the intersection of traffic,the traffic situation is complicated,and the problem of occlusion and image deformation caused by vehicle motion and shooting angle makes the highprecision and anti-interference license plate recognition system have important research value.In recent years,with the development of deep learning technology,especially in the field of image recognition and object detection,it has brought about a huge performance improvement compared with the traditional methods.In this paper,we tried to use the method of deep convolutional neural network to propose an recognition algorithm for Chinese mainland license plates,which is excellent in accuracy and robustness.The recognition accuracy of this method on our private dataset reaches 94.93%. It takes an average of 243 milliseconds to identify the picture.This method can be generalized to the license plate recognition problem in other countries by changing the data set.
Vehicle license plate recognition is an important application of computer vision technology and plays a key role in ensuring daily traffic order.Therefore,license plate recognition technology has always been a hot research in the field of computer vision.Since the application scene of the license plate recognition system is usually at the intersection of traffic,the traffic situation is complicated,and the problem of occlusion and image deformation caused by vehicle motion and shooting angle makes the highprecision and anti-interference license plate recognition system have important research value.In recent years,with the development of deep learning technology,especially in the field of image recognition and object detection,it has brought about a huge performance improvement compared with the traditional methods.In this paper,we tried to use the method of deep convolutional neural network to propose an recognition algorithm for Chinese mainland license plates,which is excellent in accuracy and robustness.The recognition accuracy of this method on our private dataset reaches 94.93%. It takes an average of 243 milliseconds to identify the picture.This method can be generalized to the license plate recognition problem in other countries by changing the data set.
引文
[1]Ioffe S,Szegzdyc.Batch Normalization:Accelerating Deep Network Training by Reducing Internal Covariate Shift[J]Computer Science,2015,3(2):1-9.
[2]He K,et al.Deep Residual Learning for Image Recognition[J].IEEE Computer Society,2016,(1):770-778.
[3]Girshick R.Fast R-CNN[M]//IEEE arxivBibliographic Explorer,2015.
[4]Ren S,He K,Girshick R,et al.Faster R-CNN:Towards Real-Time Object Detection with Region Proposal Networks[J].IEEETransactions on Pattern Ana-lysis&Machine Intelligence,2017,39(6):1137-1149.
[5]Liu W,Anguelov D,Erhan D,et al.SSD:Single Shot MultiBox Detector[J]//European Conference on Computer Vision,2016,(9905):21-37.
[6]Redmon J,Divvala S,Girshick R,et al.You Only Look Once:Unified,Real-Time Object Detection[J].2015.
[7]Redmon J,Farhadi A.YOLO9000:Better,Faster,Stronger[C]//IEEE Conference on Computer Vision and Pattern Recognition(CVPR)-Honolulu,2017:6517-6525.
[8]Redmon J,Farhadi A.YOLOv3:An Incremental Improve-ment[J].2018.
[9]Bodla N,Singh B,Chellappa R,et al.Soft-NMS——Improving Object Detection With One Line of Code[J].2017.
[10]Liu Y,Huang H,Cao J,et al.Convolutional neural networks-based intelligent recognition of Chinese license plates[J].Soft Computing,2017.
[11]B.Li,B.Tian,Q.Yao,K.Wang,A vehicle license plate recognition system based on analysis of maximally stable extremal regions,in:IEEE International Conference on Networking,Sensing and Control,2012,pp.399-404.
[12]J.Dun,S Zhang,X.Ye,Y.Zhang,Chinese license plate localization in multi-lane with complex background based on concomitant colors,IEEE Intell.Transp.Syst.Mag.7(3)(2015)51-61.
[13]Scalable object detection using deep neural networks.In:CVPR.(2014)
[14]Scalable,high-quality object detection.arXiv preprint arXiv:1412.1441 v3(2015).
[15]Chollet,Fran?ois.Xception:Deep Learning with Depthwise Separable Convolutions[J].2016.
[2]He K,et al.Deep Residual Learning for Image Recognition[J].IEEE Computer Society,2016,(1):770-778.
[3]Girshick R.Fast R-CNN[M]//IEEE arxivBibliographic Explorer,2015.
[4]Ren S,He K,Girshick R,et al.Faster R-CNN:Towards Real-Time Object Detection with Region Proposal Networks[J].IEEETransactions on Pattern Ana-lysis&Machine Intelligence,2017,39(6):1137-1149.
[5]Liu W,Anguelov D,Erhan D,et al.SSD:Single Shot MultiBox Detector[J]//European Conference on Computer Vision,2016,(9905):21-37.
[6]Redmon J,Divvala S,Girshick R,et al.You Only Look Once:Unified,Real-Time Object Detection[J].2015.
[7]Redmon J,Farhadi A.YOLO9000:Better,Faster,Stronger[C]//IEEE Conference on Computer Vision and Pattern Recognition(CVPR)-Honolulu,2017:6517-6525.
[8]Redmon J,Farhadi A.YOLOv3:An Incremental Improve-ment[J].2018.
[9]Bodla N,Singh B,Chellappa R,et al.Soft-NMS——Improving Object Detection With One Line of Code[J].2017.
[10]Liu Y,Huang H,Cao J,et al.Convolutional neural networks-based intelligent recognition of Chinese license plates[J].Soft Computing,2017.
[11]B.Li,B.Tian,Q.Yao,K.Wang,A vehicle license plate recognition system based on analysis of maximally stable extremal regions,in:IEEE International Conference on Networking,Sensing and Control,2012,pp.399-404.
[12]J.Dun,S Zhang,X.Ye,Y.Zhang,Chinese license plate localization in multi-lane with complex background based on concomitant colors,IEEE Intell.Transp.Syst.Mag.7(3)(2015)51-61.
[13]Scalable object detection using deep neural networks.In:CVPR.(2014)
[14]Scalable,high-quality object detection.arXiv preprint arXiv:1412.1441 v3(2015).
[15]Chollet,Fran?ois.Xception:Deep Learning with Depthwise Separable Convolutions[J].2016.