基于Faster R-CNN的车辆多属性识别
|
摘要
基于Faster R-CNN提出一种车辆的多属性识别模型。首先利用Faster R-CNN对车辆数据库进行训练,得到车辆检测网络,对图像中多个车辆目标进行检测。将检测结果输入改进的车辆属性识别网络中,对检测得到的车辆进行属性推断,包括车辆颜色、品牌和姿态。为评估车辆检测精度和车辆多属性识别的准确率,采集了8 000张真实场景下的图片作为测试集进行测试。对于车辆检测网络,对比了R-CNN、Fast R-CNN等方法的检测精度;对于车辆属性识别,对比了不同网络结构、不同图片分辨率和单属性和多属性等对于识别准确率的影响。实验结果表明,基于Faster R-CNN的车辆多属性识别方法充分学习了不同属性间的特征,具有较高的准确率和检测精度,以及良好的通用性和鲁棒性,适用于车辆多属性分类。
We put forward a vehicle multi-attribute recognition model based on Faster R-CNN. First we use vehicle images to train and obtain vehicle detection network for detection of vehicle targets in image. Then we put detected results into vehicle attribute recognition network and infer attributes including color, type and view. In order to evaluate the accuracy of vehicle detection precision and vehicle multi-attributes detection,we collect 8 000 vehicle images under actual scene as test set for testing. In terms of vehicle detection network,we compare the detection precision of R-CNN and Fast R-CNN, and for vehicle attribute recognition we compare the accuracy of different network,different image definition and different number of attributes. Experiment shows that the proposed vehicle multi-attribute recognition method based on Faster R-CNN can learn more features fully with higher accuracy and precision, as well as better versatility and robustness,which can be used for vehicle multi-attribute classification.
We put forward a vehicle multi-attribute recognition model based on Faster R-CNN. First we use vehicle images to train and obtain vehicle detection network for detection of vehicle targets in image. Then we put detected results into vehicle attribute recognition network and infer attributes including color, type and view. In order to evaluate the accuracy of vehicle detection precision and vehicle multi-attributes detection,we collect 8 000 vehicle images under actual scene as test set for testing. In terms of vehicle detection network,we compare the detection precision of R-CNN and Fast R-CNN, and for vehicle attribute recognition we compare the accuracy of different network,different image definition and different number of attributes. Experiment shows that the proposed vehicle multi-attribute recognition method based on Faster R-CNN can learn more features fully with higher accuracy and precision, as well as better versatility and robustness,which can be used for vehicle multi-attribute classification.
引文
[1]康维新,曹宇亭,盛卓,等.车辆的Harris与SIFT特征及车型识别[J].哈尔滨理工大学学报,2012,17(3):69-73.
[2]曹晓明.基于多图像特征金字塔的车辆检测[D].北京:北京交通大学,2016.
[3] FELZENSZWALB P,MCALLESTER D,RAMANAN D.A discriminatively trained,multiscale,deformable part model[C]//IEEE conference on computer vision and pattern recognition.Anchorage,AK,USA:IEEE,2008:1-8.
[4] GIRSHICK R,DONAHUE J,DARRELL T,et al.Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Proceedings of the IEEE conference on computer vision and pattern recognition.[s.l.]:IEEE,2014:580-587.
[5] GIRSHICK R.Fast r-cnn[C]//Proceedings of the IEEE international conference on computer vision. Santiago,Chile:IEEE,2015:1440-1448.
[6] REN Shaoqing,HE Kaiming,GIRSHICK R,et al.Faster RCNN:towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis&Machine Intelligence,2015,39(6):1137-1149.
[7] LI Xiuzhi,ZHANG Guanming,FANG Jing,et al.Vehicle color recognition using vector matching of template[C]//Third international symposium on electronic commerce and security.Guangzhou,China:IEEE,2010:189-193.
[8]张军,张婷,杨正瓴,等.深度卷积神经网络的汽车车型识别方法[J].传感器与微系统,2016,35(11):19-22.
[9]唐朝霞,俞扬信,张粤.基于粒子群算法和云模型的车型识别[J].微电子学与计算机,2013,30(11):80-83.
[10]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.
[11] SZEGEDY C,LIU Wei,JIA Yangqing,et al.Going deeper with convolutions[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. Boston,MA,USA:IEEE,2015:1-9.
[12] KRIZHEVSKY A,SUTSKEVER I,HINTON G E. ImageNet classification with deep convolutional neural networks[C]//Proceedings of the 25th international conference on neural information processing systems.Lake Tahoe,Nevada:Curran Associates Inc.,2012:1097-1105.
[13] YANG Linjie,LUO Ping,LOY C C,et al.A large-scale car dataset for fine-grained categorization and verification[C]//IEEE conference on computer vision and pattern recognition.Boston,MA,USA:IEEE,2015:3973-3981.
[14] RUSSAKOVSKY O,DENG J,SU H,et al.ImageNet large scale visual recognition challenge[J].International Journal of Computer Vision,2014,115(3):211-252.
[15] JIA Yangqing,SHELHAMER E,DONAHUE J,et al. Caffe:convolutional architecture for fast feature embedding[C]//Proceedings of the 22nd ACM international conference on multimedia.Orlando,Florida,USA:ACM,2014:675-678.
[2]曹晓明.基于多图像特征金字塔的车辆检测[D].北京:北京交通大学,2016.
[3] FELZENSZWALB P,MCALLESTER D,RAMANAN D.A discriminatively trained,multiscale,deformable part model[C]//IEEE conference on computer vision and pattern recognition.Anchorage,AK,USA:IEEE,2008:1-8.
[4] GIRSHICK R,DONAHUE J,DARRELL T,et al.Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Proceedings of the IEEE conference on computer vision and pattern recognition.[s.l.]:IEEE,2014:580-587.
[5] GIRSHICK R.Fast r-cnn[C]//Proceedings of the IEEE international conference on computer vision. Santiago,Chile:IEEE,2015:1440-1448.
[6] REN Shaoqing,HE Kaiming,GIRSHICK R,et al.Faster RCNN:towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis&Machine Intelligence,2015,39(6):1137-1149.
[7] LI Xiuzhi,ZHANG Guanming,FANG Jing,et al.Vehicle color recognition using vector matching of template[C]//Third international symposium on electronic commerce and security.Guangzhou,China:IEEE,2010:189-193.
[8]张军,张婷,杨正瓴,等.深度卷积神经网络的汽车车型识别方法[J].传感器与微系统,2016,35(11):19-22.
[9]唐朝霞,俞扬信,张粤.基于粒子群算法和云模型的车型识别[J].微电子学与计算机,2013,30(11):80-83.
[10]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.
[11] SZEGEDY C,LIU Wei,JIA Yangqing,et al.Going deeper with convolutions[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. Boston,MA,USA:IEEE,2015:1-9.
[12] KRIZHEVSKY A,SUTSKEVER I,HINTON G E. ImageNet classification with deep convolutional neural networks[C]//Proceedings of the 25th international conference on neural information processing systems.Lake Tahoe,Nevada:Curran Associates Inc.,2012:1097-1105.
[13] YANG Linjie,LUO Ping,LOY C C,et al.A large-scale car dataset for fine-grained categorization and verification[C]//IEEE conference on computer vision and pattern recognition.Boston,MA,USA:IEEE,2015:3973-3981.
[14] RUSSAKOVSKY O,DENG J,SU H,et al.ImageNet large scale visual recognition challenge[J].International Journal of Computer Vision,2014,115(3):211-252.
[15] JIA Yangqing,SHELHAMER E,DONAHUE J,et al. Caffe:convolutional architecture for fast feature embedding[C]//Proceedings of the 22nd ACM international conference on multimedia.Orlando,Florida,USA:ACM,2014:675-678.