An Approach to Evaluation of Potential Danger of Vehicles Using Video Image Data on Traffic Flow

在全面回顾国内外交通安全研究现状的基础上，指出应从以下几个方面进行研究：1)在对实际交通现象进行分析的基础上，了解引起交通冲突或事故的主要因素；2)开发合适的交通安全评价指标；3)开发预测交通安全措施效果的工程工具。鉴于此，本研究开发了一种利用从视频图像中提取的车辆轨迹数据来评价交通流安全性的方法，包括对路段上车辆运动的连续观察。以下是我们在这项研究中所做的工作。针对分析路段经常发生的交通事故类型，提出了定量评价交通冲突的方法。为了验证这些方法在交通流安全评价中的适用性，本研究分析了事故发生频率与冲突严重程度之间的关系，所提出的冲突指数衡量冲突的严重程度。利用冲突指数从车辆轨迹数据中提取潜在危险车辆的运动，结合车辆类型和大小、变道行为等信息，研究交通流中危险事件发生的主要原因和过程。本研究从微观的角度对车辆的行为进行了分析，建立了适用于车辆间距很小的密集交通条件下的跟车、变道和速度调节模型。这些模型有望被纳入微观交通仿真模型中，用于评价交通安全和道路设计。

Based on an overall review about researches to improve traffic safety, it can be said that our attentions should be paid to the following research topics : 1) understanding the major factors to cause traffic conflicts or accidents based on the analyses of actual traffic phenomena, 2) developing the suitable indices to evaluate the traffic safety, and 3) developing the engineering tools to predict the effect of traffic safety measures. Considering the above, this study develops an approach to evaluate the traffic flow safety using the vehicular trajectory data extracted from the video image including the continuous observation of vehicular movements along a road section. The followings are what we have done in this study.1. This study develops the approaches to quantitatively evaluate the traffic conflicts considering the types of accidents frequently observed on the analyzed road section. In order to confirm the applicability of these approaches for evaluating traffic flow safety, this study analyzes the relation between the frequency of accidents and the severity of conflicts measured by the conflict indices proposed.2. Focusing on the movements of potentially dangerous vehicles extracted from the vehicular trajectory data using the conflict indices, together with the information on vehicle type and size, lane changing behavior and so on, this study investigates the major causes and process of occurrence of dangerous event in the traffic flow.3. This study analyzes the vehicular behavior in a microscopic way and develops car-following, lane-changing and speed adjustment models that are applicable for the dense traffic condition with very short spacing between vehicles. It is expected that these models might be incorporated into the microscopic traffic simulation model to evaluate traffic safety and road design.