基于双视角学习原理的交通视频车辆事件鲁棒检测

提出了1种基于双视角学习原理的高速公路交通视频车辆事件鲁棒检测算法.针对道路交通结构化特点提出了分车道外极面图(Epipolar Plane Image,简称 EPI),以此反映交通断面车流整体特征.基于双视角学习原理,融合现有广泛应用的反映车辆独立行为的行驶轨迹特征,实现高速公路车辆事件鲁棒检测.针对多种典型车辆事件(包括交通拥堵,车辆逆行,车辆违规停车,交通事故等),本文算法总体检测率为94.09%,误检率为4.51%,漏检率为1.40%,其性能与传统单视角方法比较有较大的提高.

A Robust Video Detection Method for Traffic Incidents based on the Perception Theory of Dual Perspectives

A new highway incident detection algorithm is proposed based on a perception theory of dual perspectives.The epi-polar plane image (EPI)features with each separate traffic lane were developed,which described the overall characteristics of traffic in each pre-defined lane.When fused with the separated behavior of each vehicle trajectories,the higher performance as well as lower complexity with algorithm is achieved based on the integrated dual perspective perceiving framework.Extensive experimental results with variety of typical traffic incidents (such as traffic congestion,retrograde driving vehicles,vehicle parking violations,and traffic collisions)showed that this new highway incident detection algorithm has the detection rate 94.09%,the false detection rate 4.51%, and the false negative rate1.40%.