Development of a System «ViDeS» for Monitoring the Dynamics of Traffic Flows Based on the Virtual Detectors Method

Abstract

The usage of computer vision methods to analyze such characteristics of traffic flows as intensity, density and velocity is an integral part of the functioning of modern intelligent transport systems. However, during video data capturing, there are a lot of crucial factors, such as the angle and video scale of camera, road illumination level, weather conditions and etc., that directly affect the detection accuracy. All of this requires the development of software with flexible settings allowing to achieve maximum accuracy for each specific case. The paper presents a description of the system for assessing the intensity, density and velocity of traffic flows based on the technology of virtual detectors - "ViDeS". The results of the system operation based on an array of video data collected from the higher mathematics department camera are presented. The system setup options and their influence on the detection accuracy depending on the time of day are described. A study of the processing time and detection accuracy depending on the number of processed video frames is presented.