基于概率图模型的机动车逆行行为识别

Recognition of Retrograde Behavior of Motor Vehicles Based on Probability Graph Model

由于车载定位设备精度的不断提高,利用车辆的轨迹数据可以识别车辆的逆行行为,但设备或环境会产生噪声,本文主要针对轨迹处理过程中的噪声问题,研究概率图模型下的逆行行为概率判断。首先对机动车出行轨迹进行识别,拆分成出行轨迹,解决因为掉头等对概率判断的影响;再利用动态贝叶斯网络建立马尔科夫模型,对识别结果空间推断;进而基于方向角数据结合卡尔曼滤波法进行去噪后与地图匹配,计算定位点在道路上的先验概率和条件概率,再利用相邻轨迹点的条件概率计算该点的逆行概率,最终设定逆行行为判定阈值,并利用已知的轨迹数据验算准确性。结果显示,利用动态贝叶斯网络能够发现机动车的逆行行为,对提高道路安全有重要意义。

With the improvement of the accuracy of vehicle-mounted positioning equipment,the retrograde behavior of vehicles can be identified by using the trajectory data of vehicles.However,equipment or environment will produce noise,so this paper mainly focuses on the noise problem in the process of trajectory processing and studies the probability judgment of retrograde behavior under the probability graph model.In this paper,the vehicle travel trajectory is firstly identified and divided into travel trajectory,so as to solve the influence of turning around and other factors on probability judgment.Then dynamic Bayesian network is used to establish markov model and spatial inference of recognition results is made.Then the direction Angle and other data in the data are combined with kalman filtering method for denoising and map matching.The prior probability and conditional probability of the anchor point on the road are calculated,and the retrograde probability of the point is calculated by using the conditional probability of adjacent track points,and finally the threshold of retrograde behavior is set,and the accuracy of calculation is checked by using the known track data.The results show that the use of dynamic Bayesian network can detect vehicle retrograde behavior,which is of great significance to improve road safety.