基于得分系数的跟车工况驾驶风格识别研究

Based on NGSIM database, THW and ITTC were selected as parameters to evaluate the collision risk level. And a rapid recognition metric, called the score coefficient of objectivity (SCO), was proposed to measure the driving radicalness in the sampling period with a standard value between 0 and 1. Furthermore, the decision boundary of SCO was analyzed to avoid miscarriage of justice. The accuracy of the score coefficient of objectivity classification was evaluated based on K-Means clustering algorithm. The results show that, compared with traditional classification algorithms, the new method can provide a better veracity and real-time performance. The overall accuracy rate can reach up to 95.54% and the boundary miscarriage of justice can reduce to 4.46%. When the model parameters and evaluation methods are applied to new condition, the 94% overall accuracy rate can also be obtained. Based on the new method, a real-time and convenient driving style recognition system can be developed to achieve a cooperating and individuation control for the advanced driving.