AI-based Affordable High-density Traffic Monitoring System

Vehicle detection and traffic monitoring system play a significant role in many areas of transportation infrastructure. Autonomous vehicle detection and computations of the number of vehicles are becoming increasingly crucial in highway traffic management. In Bangladesh, especially in Dhaka, a substantial amount of time from daily life is wasted due to traffic congestion. However, the vehicle monitoring system is very challenging in highly dense traffic areas, and researchers are working relentlessly to improve the performance. In this paper, we have proposed an affordable AI-based vehicle detection and traffic monitoring system to fulfill this need. To validate our model, we have done experiments to detect five different vehicle types and implemented the model in Raspberry Pi. The experiment results and edge implementation indicate that our technique can lead to a simple, affordable solution for traffic monitoring and ultimately reduce traffic congestion. For still images, minimum accuracy 86%, and for traffic footage, minimum accuracy 80% is achieved by our experiment.