Towards ITS Vision Assisted Cooperative Perception

Abstract

Undoubtedly ITS technologies are maturing as an everyday reality and vehicles are being introduced as information providers on a very complex multi-agent system. Consequently, legacy vehicles, without such capabilities, will become a hindrance to the market introduction phase and consequently a liability to the overall user safety. In this paper, we aim to help on mitigating this issue by presenting a cooperative awareness system that can be easily coupled with already available systems. The system is to be used by different agents since it detects road traffic obstacles. It also provides vehicle collision avoidance to rear- ended accidents by warning the driver of the potential risk. The information gathered is then broadcasted to alert the surrounding environment. The proposed algorithm was developed using Open- CV which provides an open source library for computer vision algorithms. The system operation is divided into three steps, firstly a smart-phone camera is used for car detection. Then, the distance to the vehicles is computed and the risk of collision is evaluated. Finally, the information is broadcasted notifying other vehicles of the presence of the proximity of the vehicle. The experimental and obtained results show the efficiency of the developed system in real-world scenarios. In terms of detection accuracy, the evaluations show a 91% precision rate.