基于图像处理的信号交叉口交通冲突的检测与判定

发布时间：2018-05-16 00:00

本文选题：信号交叉口 + 直左冲突　；参考：《西南交通大学》2017年硕士论文

【摘要】：城市道路的信号交叉口对于城市安全运行至关重要。作为重要的道路节点,在交叉口处有大量的交通参与者汇集在一起,非常容易发生交通冲突与交通事故。目前得研究大多是在定性的安全评价分析,缺少事故数据,基于交通冲突技术的研究也大多是宏观的分析安全问题,多数是冲突类型、机理等。对于冲突发生与冲突严重程度的判定需要从微观上更进一步研究。针对上述问题,本文以信号交叉口的直行车与左转车的冲突(直左冲突)为研究对象进行分析。首先,对交通冲突的理论以及信号交叉口的直行车、左转车、直左冲突的交通特性进行分析,对比各种冲突判别方法的优劣,选择减速度作为判定冲突是否发生的指标。在判定发生了冲突之后,进一步利用减速度进行冲突严重程度的划分。针对传统的冲突观测方法的局限性,提出用视频图像处理技术来获得车辆的轨迹。为获取指标的有效数据,首先进行了合理的冲突观测录像,再利用Tracker软件对调查的视频图像进行直左交互作用中的车辆运动轨迹信息提取。并且对产生的误差进行分析,确保数据的精确性。划分了两车交互时的运动状态,明显冲突、干扰情况、无冲突。根据可插入间隙理论,对于干扰情况下的数据,将左转车优先通过冲突点,作为左转车的接受间隙,选择WU方法,确定冲突发生的减速度临界值。在将所有数据进行冲突发生判断后,根据减速度的判别标准,将冲突划分为三个等级,一般冲突、中等冲突、严重冲突,通过数据模拟得到指标的分布,获得第一个拐点,并且确定各个等级的临界值,验证了利用视频技术进行冲突判定的有效性,实现了图像处理的交通冲突的自动判别。在冲突判定的结果基础上,进一步探索了研究冲突的实际意义。
[Abstract]:The signalized intersections of urban roads are very important for the safe operation of cities. As an important road node, there are a large number of traffic participants at the intersection, which is prone to traffic conflicts and traffic accidents. At present, most of the researches are based on qualitative safety evaluation and analysis, and lack of accident data. The research based on traffic conflict technology is also mostly macro-analysis of safety issues, mostly conflict types, mechanisms and so on. The determination of conflict occurrence and severity needs to be further studied microscopically. In order to solve the above problems, this paper analyzes the conflict between the direct vehicle and the left-turn vehicle at the signalized intersection. Firstly, the theory of traffic conflict and the traffic characteristics of direct vehicle, left turn vehicle and straight left conflict at signalized intersection are analyzed, the advantages and disadvantages of various conflict discrimination methods are compared, and the deceleration is selected as the index to judge whether the conflict occurs or not. After the conflict is judged, the severity of the conflict is further divided by using deceleration. In view of the limitation of the traditional conflict observation method, a video image processing technique is proposed to obtain the vehicle trajectory. In order to obtain the effective data of the target, a reasonable conflict observation video is made, and then the vehicle motion track information is extracted from the investigated video image by using Tracker software in the direct to left interaction. And the error is analyzed to ensure the accuracy of the data. The motion states of the two vehicles are divided into two parts: obvious conflict, interference and no conflict. According to the theory of pluggable clearance, for the data in the case of interference, the left-turn vehicle passes through the conflict point first, as the acceptance gap of the left-turn vehicle, and chooses the Wu method to determine the critical value of the deceleration speed of the conflict. After judging the conflict of all the data, according to the criterion of deceleration, the conflict is divided into three levels: general conflict, moderate conflict, serious conflict. The distribution of indicators is obtained through data simulation, and the first inflection point is obtained. At the same time, the critical value of each level is determined, which verifies the validity of conflict determination using video technology, and realizes the automatic identification of traffic conflicts in image processing. Based on the results of conflict determination, the practical significance of the study of conflict is further explored.
【学位授予单位】：西南交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U495;TP391.41

【参考文献】