Object Tracking Method for Videomonitoring in Intelligent Transport Systems

Traffic management is one of the fundamental tasks of intelligent transport systems (ITS). It includes real-time data collection from the roads, its processing, and transmission of useful information to various decision-making systems. Loop detectors and different kinds of integrated into road infrastructure sensors can describe traffic flow for an adaptive traffic control system, city security system, etc. Vision-based technology (CCTV camera) can provide more detailed information about the traffic flow. It can identify anomalies and traffic incidents, which is very important for traffic police. This will allow us to instantly respond to situations and prevent traffic jams. An urban video monitoring system has to be supported by smart computer vision algorithms. There are many object detection and tracking algorithms that could be used in video monitoring systems. However, we faced the problem of inaccurate work of the algorithms in a real environment with occlusions, video stream interruptions. It significantly affects on object tracking, which is important for the task of vehicle counting, speed detection, and trajectory formation. In this paper, we developed and evaluated object tracking method on video streams from real road environment. We compared the results with other tracking algorithms.