摘要
针对现有车载行人预警系统预警准确度较低的问题,采用激光雷达和行车记录仪对车辆-行人的交互冲突数据进行采集,建立基于隐马尔可夫模型(Hidden Markov Model,HMM)的目标轨迹预测模型,利用Baum-Welch算法和Viterbi算法分别进行参数训练和轨迹预测,并提出一种车辆-行人冲突预警规则。验证结果表明:将车辆到达斑马线的时间等于3s时作为冲突预警阈值最为合理,此时冲突的识别率为79.6%,可以较为准确地预测出车辆和行人可能存在的冲突,为车载行人预警系统提供有效参考。
To address the problem of low warning accuracy of the existing vehicle-pedestrian warning system,the vehicle-pe⁃destrian interaction conflict data is collected by LiDAR and vehicle recorder,a target trajectory prediction model based on hidden Markov model(HMM)is established,and Baum-Welch algorithm and Viterbi algorithm are used for parameter training and the Baum-Welch algorithm and Viterbi algorithm are used for parameter training and trajectory prediction,and a vehicle-pedestrian conflict warning rule is proposed.The validation results show that it is most reasonable to use the time when the vehicle reaches the crosswalk equal to 3 seconds as the conflict warning threshold,and the conflict recognition rate is 79.6%at this time,which can pre⁃dict the possible conflicts between vehicles and pedestrians more accurately and provide an effective reference for the in-vehicle pe⁃destrian warning system.
作者
施雯
刘艳娟
赵彬
SHI Wen;LIU Yanjuan;ZHAO Bin(Department of Finance and Economics,Shaanxi Youth Vocational College,Xi'an 710068;School of Automobile,Chang'an University,Xi'an 710064)
出处
《计算机与数字工程》
2022年第7期1480-1484,1603,共6页
Computer & Digital Engineering
基金
国家自然科学基金项目(编号:51908054)
2021年度陕西省职业技术教育学会课题(编号:2021SZXYB32)资助。
关键词
车载行人预警
隐马尔可夫
轨迹预测
冲突预警
vehicular pedestrian warning
hidden Markov
trajectory prediction
conflict warning
