| 492 | 16 | 199 |
| 下载次数 | 被引频次 | 阅读次数 |
准确掌握地铁车辆内拥挤程度是提高城市轨道交通服务质量的手段之一。本文在对地铁车辆监控视频图像提取与分析的基础上,提出了一种基于卷积神经网络的车辆拥挤度识别方法。该方法使用车辆监控视频建立了车厢乘客数据集,通过提取视频图像检测区域以及人群特征检测来实现地铁列车车辆拥挤度识别。实验结果表明,所提出的方法检测速度快,能够满足实际应用中实时性要求,三级拥挤度分类识别实验准确度为98%,四级拥挤度分类识别实验准确度为87%,其检测结果可辅助城市轨道交通管理者快速掌握线网实时客流拥挤情况。
Abstract：Accurately analyzing and obtaining the congestion levels in subway trains is one way of improving the quality of urban rail transit services. This paper proposes a vehicle congestion recognition method using a convolutional neural network to extract and analyze subway vehicle monitoring video images. This method uses the vehicle monitoring video to build a dataset of passengers in the carriage and uses a convolutional neural network with a cascade structure to realize recognition of subway train vehicle congestion by extracting the detection area of the video image and detecting crowd characteristics. The experiment results demonstrated that our proposed method had a fast detection speed and can meet real-time requirements in practical applications. The accuracy of three-level congestion degree experiment is 98%, and the accuracy of four-level congestion degree experiment is 87%. The test results can assist urban rail transit managers to quickly understand real-time passenger flow congestion in the line network.
[1]中华人民共和国建设部.城市轨道交通工程项目建设标准[M].北京:中国计划出版社,2008.
[2]北京市质量技术监督局.城市轨道交通工程设计规范[M].北京:北京地方标准,2013.
[3]卢恺,韩宝明,鲁放.城市轨道交通运营客流数据分析缺陷及应对[J].都市快轨交通,2014,27(4):25-29.
[4]魏梦.基于卷积神经网络的人群密度分析[D].合肥:中国科学技术大学,2018.
[5]沈娜.智能视频监控系统中人群密度估计与实现[D].南京:南京航空航天大学,2012.
[6]陶茂辉.基于视频图像的人群密度估计研究[D].成都:电子科技大学,2014.
[7]ZHANG C,LI H,WANG X,et al.Cross-scene crowd counting via deep convolutional neural networks[C]//Conference on Computer Vision and Pattern Recognition.Boston:IEEE press,2015:833-841.
[8]ZHANG Y Y.Single-image crowd counting via multicolumn convolutional neural network[C]//Conference on Computer Vision and Pattern Recognition.Las Vegas:IEEE press,2016:589-597.
[9]SAM D B,SURYA S.Switching convolutional neural network for crowd counting[C]//Conference on Computer Vision and Pattern Recognition.Hawaii:IEEE press,2017:429-437.
[10]SINDAGI V A,PATEL V M.Cnn-based cascaded multitask learning of high-level prior and density estimation for crowd counting[C]//Conference on Advanced Video and Signal Based Surveillance.Piscataway:IEEE press,2017:1-6.
[11]袁烨,吴震宇,江先志.基于卷积神经网络的人群计数研究[J].成组技术与生产现代化,2017,34(2):49-53.
[12]张霖,韩宝明,李得伟.基于图像技术的城市轨道交通大客流辨识[J].都市快轨交通,2012,25(1):72-77.
[13]刘旭.视频监控中的目标计数方法研究[D].合肥：中国科学技术大学,2018.
[14]马海军,王文中,翟素兰,等.基于卷积神经网络的监控视频人数统计算法[J].安徽大学学报:自然科学版,2016,40(3):22-28.
[15]覃勋辉,王修飞,周曦,等.多种人群密度场景下的人群计数[J].中国图象图形学报,2013,18(4):392-398.
[16]HE K,ZHANG X,REN S,et al.Spatial pyramid pooling in deep convolutional networks for visual recognition[C]//European Conference on Computer Vision Computer Vision.Zurich:Springer press,2014:346-361.
[17]时增林,叶阳东,吴云鹏,等.基于序的空间金字塔池化网络的人群计数方法[J].自动化学报,2016,42 (6):866-874.
[18]马海军.监控场景中人数统计算法的研究与应用[D].合肥：安徽大学,2016.
[19]SINDAGI V A.CNN-based cascaded multi-task learning of high-level prior and density estimation for crowd counting[EB/OL].https://github.com/svishwa/crowdcountcascaded-mtl,2017.
基本信息:
中图分类号:U29-39;TP391.41
引用信息:
[1]张杏蔓,鲁工圆.基于视频图像分析的地铁列车车辆拥挤度识别方法研究[J],2020,18(03):142-152.
基金信息:
国家重点研发计划(2017YFB1200701);; 四川省教育厅2018—2020年高等教育人才培养质量和教学改革项目(JG2018-135)