Traffic Surveillance to detect wrong-way vehicles by image processing

Abstract

Wrong-way driving is one of the throughout the process causes of road accidents and traffic blocks all over the worldwide. By detecting wrong-way vehicles, the frequency of accidents may be minimized and traffic congestion can be alleviated. With the rising popularity of real time traffic management systems and thanks to the availability of cheaper cameras, the surveillance video has become a big source of data. In this study, we present an autonomous wrong-way vehicle identification system using on-road security camera footage. Our system operates in three stages: the recognition of vehicles from the video frame by utilizing the You Only Look Once (YOLO) method, track each vehicle in a specific zone of interest using the centroid tracking technique, and identify the wrong-way of driving vehicles. YOLO is incredibly precise in object detection and the centroid tracking method can track any moving item efficiently. Experimenting with several traffic recordings indicates that our proposed system can recognize and identify any wrong-way vehicle in varied light and weather situations. The system is very simple and straightforward to apply.