Evaluation Method for Aggressiveness of Driving Behavior Using Drive Recorders
Evaluation Method for Aggressiveness of Driving Behavior Using Drive Recorders
复制标题
DOI：
10.1541/ieejjia.4.59
复制
发表时间：
2015-01-01
影响因子：
1.7
通讯作者：
Takeda, Kazuya
中科院分区：
文献类型：
--
作者：
Li, Yiyang;Miyajima, Chiyomi;Takeda, Kazuya
在本文中，我们提出了一种通过使用行车记录仪的驾驶信号来测量驾驶行为的攻击性的自动化方法。目前，一些风险咨询公司让专家审查记录的驾驶行为，然后对驾驶员进行经验评分。然而，这种方法既耗时又昂贵，因此需要一种有效的自动驾驶员评估方法。我们假设可以通过关注四种类型的车辆操作行为来确定驾驶员行为的攻击性：转向、加速、减速以及加减速交替。攻击性得分可以分配给每个行为，然后通过使用多元线性回归整合这些行为来估计驾驶员的总体攻击性得分。我们评估了 78 名驾驶员的攻击性，并将我们的评估结果与风险咨询专家根据经验分配的攻击性分数进行了比较。经风险咨询专家评估，该方法的等级相关系数为0.74。
In this paper, we propose an automated method for measuring the aggressiveness of driving behavior by using driving signals from drive recorders. Currently, some risk consulting companies have experts review recorded driving behavior and then rate the drivers empirically. This approach is time-consuming and expensive, however, so an effective automated driver evaluation method is desired. We assumed that the aggressiveness of a driver's behavior can be determined by focusing on four types of vehicle operation behavior: steering, acceleration, deceleration, and alternation between acceleration and deceleration. Aggressiveness scores can be assigned to each of these behaviors, and a driver's overall aggressiveness score is then estimated by integrating these behaviors using multiple linear regression. We assessed the aggressiveness of 78 drivers and compared our assessments to the aggressiveness scores assigned empirically by the risk consulting experts. The proposed method achieved a rank correlation coefficient of 0.74 with the evaluations of the risk consulting experts.