基于视频检测的交通冲突判别方法和模型

摘要

为了克服交通冲突传统人工判别方法的缺陷,提出了基于视频检测的交通冲突发生判别的基本流程,将交通冲突发生的判别划分为交通参数的视频检测、坐标变换和交通冲突判别3个阶段.在借助OpenCV和Matlab标定工具箱进行交通实体检测和坐标变换的基础上,建立了以冲突角度、速度和距离等指标为核心的交通冲突综合判别模型,并通过现实场景验证该模型的有效性.%In order to overcome the defects of manual methods for classifying traffic conflicts, a new approach based on video detection is presented in this paper. The classification process for traffic conflicts is divided into the following three phases; the video detection, transformation of coordinates and classification of traffic conflict. Through using OpenCV and Matlab calibration toolbox for traffic entity detection and transformation of coordinates, a comprehensive traffic conflict classification model is established, which considers such key variables as impact angle, speed and distance. The effectiveness of the model is validated through the tests in a real scene.