摘要
为提高车站客流统计的精度使其可以准确预警,针对传统的客流统计方法步骤繁琐、准确率低等局限性,对基于卷积神经网络的行人人头检测方法进行研究。首先在常州某车站安检站台处通过高位摄像头采集行人的人头数据库;然后通过不同的行人特征提取网络(Inception模块、Resnet、Mobilenet)与Faster R-CNN、SSD、R-FCN等目标检测结构组合的方式来对比探究各种行人人头检测组合模型的准确率和检测速度,并选择最优的行人人头检测方法;最后通过模型试验分析,结果显示Inception-V2特征提取网络与Faster R-CNN目标检测结构组合的行人人头检测模型具有较高的准确率和较优的实时性,这种行人人头检测方法对客流预警具有重要的意义。
In order to improve the precision of station passenger flow statistics for accurate early-warning,this paper studies the pedestrian head detection method based on convolutional neural network which works without the limitations of traditional passenger flow statistics methods such as cumbersome procedures and low accuracy.Firstly,the paper uses a high-level camera to collect the database of pedestrian heads at a station security checkpoint in Changzhou.Then,the paper combines the feature extractors(Inception module,Resnet,Mobilenet)with the target detection networks(Faster R-CNN,SSD,and R-FCN).Finally,the paper compares the accuracy and speed of them,and chooses the optimal pedestrian head detection method.The experimental results show that the detection accuracy and instantaneity of the pedestrian detection model which combines the Inception-V2 feature extractors with Faster R-CNN networks is better.The pedestrian-headed detection method is of great significance to the passenger flow warning.
引文
[1]何鹏.基于类圆图形识别的客流量统计系统研究[D].武汉:武汉科技大学,2015.
[2]陈亮.基于光流法的视频分析技术在地铁客流监测中的应用[J].铁路通信信号工程技术,2013,10(1):58-62.
[3]Szegedy C,Toshev A,Erhan D.Deep Neural Networks for object detection[J].Advances in Neural Information Processing Systems,2013,26:2553-2561.
[4]Ba L J,Frey B.Adaptive dropout for training deep neural networks[C]//International Conference on Neural Information Processing Systems.New York:Curran Associates Inc.,2013:3084-3092.
[5]Szegedy C,Liu W,Jia Y,et al.Going deeper with convolutions[C]//Computer Vision and Pattern Recognition.Piscataway,NJ:IEEE,2015:1-9.
[6]Lei D,Peng J,Jing P,et al.GXNOR-Net:Training deep neural networks with ternary weights and activations without full-precision memory under a unified discretization framework[J].Neural Netw,2018,100:49-58.
[7]Szegedy C,Vanhoucke V,Ioffe S,et al.Rethinking the inception architecture for computer vision[C]//IEEE Conference on Computer Vision and Pattern Recognition.Washington:IEEE Computer Society,2016:2818-2826.
[8]Karpathy A,Toderici G,Shetty S,et al.Large-scale video classification with convolutional neural networks[C]//Computer Vision and Pattern Recognition.Piscataway,NJ:IEEE,2014:1725-1732.
[9]He K,Zhang X,Ren S,et al.Identity mappings in deep residual networks[C]//European Conference on Computer Vision.Berlin:Springer,Cham,2016:630-645.
[10]Russakovsky O,Deng J,Su H,et al.Imagenet large scale visual recognition challenge[J].International Journal of Computer Vision,2015,115(3):211-252.
[11]Ren S,He K,Girshick R,et al.Faster R-CNN:Towards realtime object detection with region proposal networks[J].IEEETransactions on Pattern Analysis&Machine Intelligence,2017,39(6):1137-1149.
[12]Zhang D,Zhang H,Li H,et al.RR-FCN:Rotational regionbased fully convolutional networks for object detection[C]//International Conference on Engineering Applications of Neural Networks.Berlin:Springer,Cham,2018:58-70.
[13]蔡汉明,赵振兴,韩露,等.基于SSD网络模型的多目标检测算法[J].机电工程,2017,34(6):685-688.