Identifying and Counting Vehicles in Multiple Lanes by Using a Low-Cost Vehicle-Mounted Sensor for Intelligent Traffic Management Systems

There is evidence that accessing online traffic data is a key factor to facilitate intelligent traffic management, especially at intersections. With the advent of autonomous vehicles (AVs), new options for collecting such data appear. To date, much research has been performed on machine learning to provide safe motion planning and to control modern vehicles such as AVs. However, few studies have considered using the sensing features of these types of vehicles to collect traffic information of the surrounding environment. In this study, we developed new algorithms to improve a traffic management system when the traffic is a mixture of human-driven vehicles (HDVs) and modern vehicles with different levels of autonomy. The goal is to utilize the sensing ability of modern vehicles to collect traffic data. As many modern vehicles are equipped with vehicle-mounted sensors by default, they can use them to collect traffic data. Our algorithms can detect vehicles, identify their type, determine the lane they are in, and count the number of detected vehicles per lane by considering multi-lane scenarios. To evaluate our proposed approach, we used a vehicle-mounted monocular camera. The experimental work presented here provides one of the first investigations to extract real traffic data from multiple lanes using a vehicle-mounted camera. The results indicate that the algorithms can identify the detected vehicle’s type in the studied scenarios with an accuracy of 95.21%. The accuracy of identifying the lane the detected vehicle is in is determined by two proposed approaches, which have accuracies of 91.01% and 91.73%.