摘 要：2020 年，中国的汽车保有量已经突破3.7 亿辆，基于非智能传感器辅助的人工管理模式受到的挑战越来越大。基于此，许多城市开始建设自己的智能交通系统，在这些工作中，对一些复杂场景，如十字路口，的建模是一个绕不开的难题。文章提出了一种通过分析车辆行驶轨迹等信息进行建模的新方案，与以往基于语义分割网络的建模方案相比，该方案具有稳定性更高、适用性更广、需要配置更低、速度更快的优势。
关键词：场景建模；检测跟踪；K-Means 聚类；曲线拟合
中图分类号：月的TP391.9 文献标识码：A 文章编号：2096-4706（2020）24-0089-04
Intersection Modeling Scheme Based on Behavior Analysis
QIU Qi，LI Jie
（University of South China，Hengyang 421001，China）
Abstract：In 2020，China’s car ownership has exceeded 370 million，and the artificial management mode based on nonintelligent sensor is facing more and more challenges. Based on this，many cities have started to build their own intelligent transportation systems. In these works，the modeling of some complex scenes，such as intersections，is a difficult problem. In this paper，a new modeling scheme is proposed by analyzing the vehicle trajectory and other information. Compared with the previous modeling scheme based on semantic segmentation network，this scheme has the advantages of higher stability，wider applicability，lower configuration and faster speed.
Keywords：scene modeling；detection and tracking；K-Means clustering；curve fitting
基金项目：湖南省创新创业训练计划（S202010555016）
参考文献：
[1] LV Z H ，ZHANG S B，XIU W Q. Solving the SecurityProblem of Intelligent Transportation System With Deep Learning [J/OL].IEEE Transactions on Intelligent Transportation Systems，2020（99）：1-10.[2020-03-20].https://ieeexplore.ieee.org/document/9043888.DOI:10.1109/TITS.2020.2980864
[2] WAN S H，XU X L，WANG T，et al. An Intelligent VideoAnalysis Method for Abnormal Event Detection in Intelligent TransportationSystems [J/OL].IEEE Transactions on Intelligent Transportation Systems，2020：1-9.[2020-09-09].https://ieeexplore.ieee.org/document/9190063.DOI:10.1109/TITS.2020.3017505.
[3] HAYDARI A，YILMAZ Y. Deep Reinforcement Learningfor Intelligent Transportation Systems：A Survey [J/OL].arXiv：2005.00935 [cs.LG].（2020-05-02）.https://arxiv.org/abs/2005.00935v1.
[4] BOCHKOVSKIY A，WANG C Y，LIAO H Y M. YOLOv4：Optimal Speed and Accuracy of Object Detection [J/OL].arXiv：2004.10934 [cs.CV].（2020-04-23）.https://arxiv.org/abs/2004.10934.
[5] WOJKE N，BEWLEY A，PAULUS D. Simple Online andRealtime Tracking with a Deep Association Metric [C]//2017 IEEEInternational Conference on Image Processing.Beijing：IEEE，2017：3645-3649.
作者简介：
邱祺（1999—），男，汉族，湖南益阳人，本科在读，研究方向：计算机视觉、群体智能；
李杰（2000—），男，汉族，湖南娄底人，本科在读，研究方向：软件工程、计算机视觉。