Approach to Obtaining Traffic Volume and Speed Based on Video-Extracted Trajectories

Intersections are key nodes of urban roads and constraints of traffic network capacity. And traffic data is the basis for intersection design, analysis, and management. The current measurement of vehicle speed is to use radar guns which could only measure point speed and volume measurement is manual counting which are time-and-energy-consuming. This paper introduces a method that collects the vehicle trajectories from videos taken by an unmanned aerial vehicle (UAV), with further processing of the trajectories to obtain useful traffic data such as intersection flow and vehicle speed. The program uses background subtraction and morphological operation for vehicle detection and uses KCF trackers to track vehicle targets. After extracting the trajectories of each vehicle in the field of view, it performs turning identification, counts the traffic flow of each entrance, and extracts more data such as vehicle speed and time headway. In order to assess the feasibility of the video processing method, case studies were performed based on the field data collected by a UAV at urban intersections in Nanjing, China. The results of these experiments proved that the approach can be applied to the efficient analysis of intersection volume, speed and other parameters based on UAV videos.