基于轨迹数据的营运车辆不良驾驶行为辨识
Identification of bad driving behaviors of operating vehicles based on trajectory data
朱兴林 1丁双伟 1姚亮 1刘泓君1
作者信息
- 1. 新疆农业大学交通与物流工程学院,新疆乌鲁木齐 830052
- 折叠
摘要
为精准识别不良驾驶行为驾驶员,基于粒子群(PSO)与支持向量机(SVM)提出一种组成模型PSO-SVM,用于辨识营运车辆的不良驾驶行为.首先,对预处理后的GPS轨迹数据利用大样本统计方法提取驾驶行为特征指标参数构建驾驶行为数据集;基于四分位差法与CRITIC赋权法确定驾驶行为特征指标参数阈值和权重,建立不良驾驶行为风险评估模型,并根据驾驶员风险评估等级划分对驾驶员的整体行驶状态是否危险进行区分.其次,通过粒子群算法(PSO)和遗传算法(GA)分别对模型的关键参数进行寻优,来确定模型的最优参数组合.最后,以驾驶行为特征指标参数作为输入指标,聚类标签作为模型的识别输出指标.结果表明:基于PSO优化后的参数C为6.402 4,γ为0.1;PSO-SVM模型对不良驾驶行为识别的精确率为91.17％、召回率为88.57％,且此模型的平衡分数F1值最高为0.89.可见PSO-SVM模型相比于普通模型对营运车辆驾驶行为的识别具有良好的性能.研究结果可为道路交通安全、营运车辆驾驶员的管理提供技术支持.
Abstract
To accurately identify bad driving behavior of drivers,a PSO-SVM model based on particle swarm optimization(PSO)and support vector machines(SVM)is proposed for identifying bad driving be-havior of commercial vehicles.Firstly,the driving behavior feature indicator parameters are extracted using a large sample statistical method from the preprocessed GPS trajectory data to construct a driving behavior dataset.Then,based on the quartile difference method and CRITIC weighting method,the threshold and weight of the driving behavior feature indicator parameters are determined to establish a bad driving behav-ior risk assessment model,and the overall driving status of the driver is distinguished according to the driv-er risk assessment level.Secondly,the key parameters of the model are optimized by PSO and genetic algo-rithm(GA)respectively to determine the optimal parameter combination of the model.Lastly,the driving behavior feature indicator parameters are used as input indicators,and the clustering label is used as the recognition output indicator of the model.The results show that the PSO optimized parameters C and y for the model are 6.402 4 and 0.1,respectively.The precision and recall rates of the PSO-SVM model for identifying bad driving behavior are 91.17％and 88.57％,respectively,and the balance score F1 value of this model is highest at 0.89.It can be seen that the PSO-SVM model has good performance in identifying driv-ing behavior of commercial vehicles compared to ordinary models.The findings of this research can offer technical assistance for enhancing road traffic safety and managing drivers of commercial vehicles.
关键词
不良驾驶行为/粒子群算法/支持向量机/交通安全Key words
Bad driving behavior/Particle swarm optimization/Support vector machines/Traffic safety引用本文复制引用
基金项目
中国学位与研究生教育学会项目(2020MSA274)
出版年
2024
NETL
NSTL
维普
万方数据