Online Identification of Abnormal Events Based on Vehicle Spatiotemporal Trajectory Characteristics
Shu De Du, Weiwei Yu · Advances in transdisciplinary engineering · 2025
Traffic incidents such as road abandoned object equips the characteristics of low probability of occurrence, high potential risks, and difficulty in inspections. Find abnormalities through manual inspections costs much as well as a not timely and comprehensive detection, so there is an urgent need for a set of methods that can detect various types of incidents. Vehicle trajectory contains rich information that reflecting the operating status of the traffic flow. Based on the emerging machine vision technology, this research collected vehicle trajectories to realize the judgment and recognition of incidents from the surveillance video. This research analyzes the traffic parameter characteristics of abnormal events and summarizes the performance of traffic parameters under different abnormal events. The road cell unit is delineated for the continuous road, and the real-time traffic parameter changes in each cell are calculated based on the vehicle trajectory. This method determines whether there are abnormalities and types of abnormal events in the road area by combining multiple parameter indicators. During the research process, the performance of traffic parameters during the occurrence of abnormal events was analyzed with specific trajectory examples, and the system is tested through nine different actual scenarios. The accuracy of incidents judgment is more than 70%.