基于视频检测的交通冲突判别方法和模型  

A Video Detection-based Traffic Conflict Classification Model

作　　者：曾炎盛[1] 马寿峰[1] 钟石泉[1] 贾宁[1] 

机构地区：[1]天津大学系统工程研究所,天津300072

出　　处：《交通信息与安全》2012年第4期11-14,61,共5页Journal of Transport Information and Safety

基　　金：国家自然科学基金项目(批准号:50908155;70971094);教育部博士点新教师基金项目(批准号:20090032120032)资助

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.

关 键 词：交通冲突判别 视频检测 冲突角度 OPENCV Matlab标定工具箱 

分 类 号：U491.265[交通运输工程—交通运输规划与管理]