危险驾驶行为辨识算法研究

Research on identification algorithm of dangerous driving behavior

为了提高危险驾驶行为辨识效能,提出利用方差贝叶斯网络模型辨识危险驾驶行为。使用车载CCD实时采集路面图像,根据自车参数确定车辆左右前轮距车道线距离,结合探头安装角度和高度确定前后车间距,通过判定一段时长内的车间距和车道线距离的方差变化确定方差模型输出;方差模型输出作为贝叶斯模型输入,综合方差模型输出与贝叶斯网络预测结果,最终判定当前驾驶行为是否存在危险。实验结果表明,所建模型能有效辨识两种危险驾驶行为,且比单一方差模型表现出更好的泛化性能。

To improve the efficiency of dangerous driving behavior identification, an algorithm to identify dangerous driving be havior by using variance bayesian network is proposed. Vehicle-mounted CCD is used to collect the real-time digital road image and the host vehicle parameters are used to determine the distance between the front wheels of the vehicle and the lane lines. Combining with the installation angle and the height of the vehicle-mounted cameras, the distance between the host vehicle and the leading vehicle is determined. The variance changes of the lane line distance and the space are used to determine the variance model output. The output of variance model is taken as the input of bayesian network model. The comprehensive result of the variance model output and bayesian network model is employed to judge whether the current driving behavior exists danger at last. The test results show that the model can effectively identify the two kinds of dangerous driving behavior and has stronger generalization performance than a single variance model.