城市快速路交通事件自动检测算法

摘要

In order to improve the accuracy of traffic incident detection for urban expressway,through analyzing the change rules of traffic flow parameters,the initial variables set of traffic incident detection which contains 12 variables was built,and the random forest method was used to select the key variables.Then combined kernel function,relevance vector machine model was constructed based on particle swarm optimization.Finally,validation and comparative analysis were carried out using inductive loop parameters measured from the north-south viaduct in Shanghai.The results show that the key variable selection can effectively improve the accuracy of traffic incident detection.The detection performance of combined kernel function RVM model is also better than that of the single kernel function RVM model and SVM model.%为了进一步提高城市快速路交通事件检测的精度,在分析交通事件上、下游交通流参数变化规律的基础上,构建包含12个变量的交通事件检测初始变量集,并采用随机森林方法对初始变量集的关键变量进行筛选,进而构建基于粒子群优化的组合核函数相关向量机模型.最后,利用上海市南北高架快速路的感应线圈实测参数进行实验验证和对比分析.研究结果表明:关键变量筛选可以有效提高交通事件检测的精度,组合核函数相关向量机模型也明显优于单一核函数相关向量机模型和支持向量机模型.