基于高精度轨迹的交叉口急变速行为识别研究

Research on Recognition of Sudden Speed Change Behavior at Intersection Based on High Precision Trajectory

文中提出通过精度可达厘米级、采样频率可达10 Hz的高精度北斗卫星定位设备获取车辆轨迹,设计车辆在交叉口前急变速行为的识别算法,采用K-Means算法对驾驶员在交叉口的急变速行为进行聚类分析,通过支持向量机进行驾驶员类型识别.在武汉市选取了1条路线,选取40名不同性别、年龄、驾龄的驾驶员驾驶试验车,通过聚类分析,将其在交叉口的驾驶类型定义为激进型和稳妥型,急变速发生位置越靠近交叉口停止线、加速度绝对值越大、发生次数越多、累计急变速时间越长的驾驶员越激进,使用支持向量机取7/10样本进行分类训练,用剩余样本进行验证,得到分类器精确率为0.83.

The vehicle trajectory was obtained by the high-precision Beidou satellite positioning equipment with the accuracy of centimeter level and the sampling frequency of 10Hz, and the identification algorithm of the vehicle’s sudden shift behavior before the intersection was designed. K-Means algorithm was used to cluster the driver’s sudden change behavior at the intersection, and support vector machine is used to identify the driver’s type. A route was selected in Wuhan, and 40 drivers with different gender, age and driving experience were selected to drive the test vehicle. Through cluster analysis, their driving types at intersections were defined as aggressive and safe. The driver who is closer to the intersection stop line, the greater the absolute value of acceleration, the more times it occurs and the longer the cumulative sudden change time is, the more radical it is. 7/10 samples are selected by support vector machine for classification training, and the remaining samples are used for verification, and the accuracy rate of the classifier is 0.83.