前方车辆驾驶行为的识别方法研究

Research on Recognition Method of the Preceding Vehicle Driving Behavior

为了降低前方车辆驾驶行为的随机变化对自身车辆安全的影响,提出了一种基于隐马尔科夫模型(Hidden Markov Model,HMM)的前方车辆驾驶行为的识别方法。首先,分析了高速公路场景中车辆的换道特性,选取前方车辆的纵向速度、横向位移和横向速度作为特征参数;采用线性均值滤波的方法对原始观测数据进行平滑滤波处理。其次,基于Baum-Welch算法训练得到前方车辆三种驾驶行为的最优隐马尔科夫模型,基于Viterbi算法识别出前方车辆的驾驶行为。最后,测试结果表明隐马尔科夫模型可以有效地识别出前方车辆的驾驶行为。

In order to reduce the effect of random variation of driving behavior of the preceding vehicle on the safety of own vehicle,this study propose a method to recognize the driving behavior of the preceding vehicle based on Hidden Markov Model. Firstly,analyze vehicle lane-changing characteristics in highway scenes,and select the longitudinal velocity,lateral displacement and lateral velocity of the preceding vehicle as characteristic parameters. And,use the linear mean filter method to smooth the original observation data. Secondly,the optimal Hidden Markov Model parameters of the three driving behavior modes of preceding vehicle are obtained based on Baum-Welch algorithm,and the driving behavior of the preceding vehicle is recognized based on Viterbi algorithm. Finally,the test results show that the Hidden Markov Model can effectively recognize the driving behavior of the preceding vehicle.