城市道路车辆行为识别方法研究  

Research on Vehicle Behavior Recognition Methods for Urban Roads

作　　者：陆金辉 鲍楠 胡晗[1] 左加阔[1] 师晓晔 潘甦[1] LU Jinhui;BAO Nan;HU Han;ZUO Jiakuo;SHI Xiaoye;PAN Su(Nanjing University of Posts and Telecommunications,Nanjing 210003,China)

机构地区：[1]南京邮电大学,江苏南京210003

出　　处：《移动通信》2023年第10期38-43,共6页Mobile Communications

基　　金：国家自然科学基金委员会面上项目“基于空间资源压缩概念和方法的物联网资源优化分配”(62071244)、“基于多域信息联合利用的空地融合频谱共享研究”(62071250);中国博士后科学基金会纵向项目“面向6G的高能效RIS-NOMA技术研究”(2021M693699)。

摘　　要：为了提高城市道路驾驶的安全性,提出一种基于混合高斯隐马尔科夫模型的城市道路车辆行为识别方法。首先通过仿真平台获取城市三车道道路场景的车辆运动轨迹信息和车辆信息并对其进行数据处理和特征提取,选取目标车辆车头中心与初始所在车道右侧车道线的距离,以及目标车辆纵向和横向速度作为特征参数,再通过鲍姆-韦尔奇算法对模型参数进行训练迭代更新,最后结合前后向算法实现城市道路车辆行为的识别。仿真结果表明,采用混合高斯隐马尔科夫模型对城市道路车辆行为进行识别具有较高的精度。In order to improve the driving safety on urban roads,a vehicle behavior recognition method is proposed based on mixed Gaussian hidden Markov model for urban roads.Firstly,we obtain the vehicle trajectory information and vehicle information of the urban three-lane road scenario through the simulation platform.The data processing and feature extraction are via selecting the distance between the center of the target vehicle's front end and the right lane line of the initial lane and taking the longitudinal speed and transverse speed of the target vehicle as the feature parameters.Then the model parameters are trained and iteratively updated through the Baum-Welch algorithm.Finally,combining the forward and backward algorithms,the recognition of the vehicle behavior is realized for urban roads.The simulation results show that the mixed Gaussian hidden Markov model is used to recognize the vehicle behavior on urban roads with high accuracy.

关 键 词：混合高斯隐马尔科夫模型 车辆行为识别 鲍姆-韦尔奇算法 前后向算法 

分 类 号：TN929.5[电子电信—通信与信息系统]