基于优化SVM的高速公路交通事件检测

为了降低交通事件对高速公路交通运行的影响,提出一种基于视频交通流参数的支持向量机交通事件检测算法。针对线圈检测器在车辆拥堵时的检测精度大幅降低,甚至无法检测等不足,采用基于视频的交通流参数检测方法。通过安装的视频摄像机,实现对高速公路全程的无盲区交通事件检测。同时,为了解决支持向量机模型参数的大小对分类器泛化能力的不利影响,采用改进的网格搜索算法对模型参数进行优化选择。应用实际采集的交通流数据,对优化的支持向量机事件检测算法进行验证,并与遗传算法、粒子群算法的优化结果进行比较。实验结果表明,提出的方法具有最高的检测率和较低的误报警率,且优化时间短,满足了交通事件检测实时性的要求。

Freeway traffic incident detection based on an optimized SVM model

To decrease the negative impact of traffic incidents on the freeway traffic operation,a traffic incident detection algorithm using a SVM model based on video-based traffic parameters is proposed. Aiming at the deficiencies of loop detectors,such as its lower detection accuracy at the time of traffic congestion, which is even impossible to detect, a video-based traffic parameter extraction method is used. With the equipped video cameras, realizing the traffic incident detection on the whole freeway without blind zones is possible. Moreover, in order to overcome the adverse effects of the parameters choosing and setting of the SVM model on the generalization of the classifier, an improved grid search algorithm is used to optimize the parameters of the SVM model. Tested by the traffic data from real video, the traffic incident detection algorithm based on the optimized SVM model is validated,and it is compared with the results of the SVM models optimized by a genetic algorithm and a particle swarm optimization algorithm. The results show that the proposed method not only has the highest detection rate and the lower false alarm rate, but also has the shortest optimizing time which can satisfy the requirement of real-time traffic incident detection.