基于Seq2Seq自编码器模型的交通事故实时检测与评价
CSTR:
作者:
作者单位:

1. 上海海事大学 经济管理学院,上海 201306;2. 朗坤智慧科技股份有限公司,南京 211100;3. 上海电科智能系统股份有限公司,上海 200063

作者简介:

通讯作者:

E-mail: jyajian@163.com.

中图分类号:

C934

基金项目:

国家自然科学基金项目(71971135,71571166).


Real-time traffic accident detection and evaluation based on Seq2Seq and autoencode model
Author:
Affiliation:

1. School of Economics and Management,Shanghai Maritime University,Shanghai 201306,China;2. Luculent Smart Technologies Co., Ltd,Nanjing 211100,China;3. Shanghai SEARI Intelligent System Co., Ltd,Shanghai 200063,China

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    路侧检测设备可以精准获取交通流量和速度等实时数据,交管部门可以借此显著提升对交通异常状态的感知水平.通过分析交通状态和交通流数据特征,建立一套基于交通流序列数据的交通事故实时检测系统和预警流程.首先,在交通状态感知方面,所建立的Seq2Seq自编码模型引入Attention机制,实现对交通状态重要特征的捕捉;其次,在交通状态异常判定方面,利用Seq2Seq自编码器对输入的原始序列数据进行重构,对比原始数据可得到结构重构误差,根据设定的阈值实现交通预警等级的判定和交通事故的实时检测;最后,以上海市延安高架的流量和速度数据为基础,分别确定不同时空状态下的事故判定阈值,并通过混淆矩阵评价方法论证所提出交通事故实时检测模型的可行性.

    Abstract:

    Roadside monitoring equipments can obtain traffic flow and speed real-time data more accurately, so that traffic management departments can significantly improve the perception of traffic anomalies. By analyzing the characteristics of traffic state and traffic flow data, this paper establishes a set of real-time traffic accident detection system and early warning process based on traffic flow sequence data. First of all, in terms of traffic state perception, this paper introduces the Attention mechanism to capture important traffic condition features based on the Seq2Seq model. Secondly, in terms of anomaly detection of traffic condition, the auto-encoder is used to realize the reconstruction of input sequence data, and the structural reconstruction error is obtained by comparing the original data, and then the real-time detection of traffic accidents and the classification of accident warning levels are realized according to the set threshold. Finally, based on the traffic and speed data of the Shanghai Yanán elevated road, the thresholds for evaluating traffic accidents in different time and space conditions are determined respectively, and the feasibility of the real-time traffic accident detection model is demonstrated by the confusion matrix evaluation method.

    参考文献
    相似文献
    引证文献
引用本文

赵超,谢天,辛国容,等.基于Seq2Seq自编码器模型的交通事故实时检测与评价[J].控制与决策,2022,37(8):2141-2148

复制
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:
  • 最后修改日期:
  • 录用日期:
  • 在线发布日期: 2022-06-29
  • 出版日期: 2022-08-20
文章二维码