Using Vehicle-Mounted Camera to Collect Information for Managing Mixed Traffic

Abstract

With increasingly rapid advances in the field of producing modern and autonomous vehicles, the need for intelligent traffic management systems, which take advantage of the vehicle's abilities to sense and communicate, has increased. A considerable amount of literature has been published on managing traffic that includes only autonomous vehicles. However, changing all vehicles to autonomous versions is a longterm process. In the near future, traffic will be a mixture of human-driven and autonomous vehicles. To date, few studies have investigated mixed traffic in intelligent management systems. The main objective of this research is to study the possibility of using a vehicle-mounted camera to sense and collect the required traffic data of the surrounding vehicles in mixed traffic. To achieve this, a vehicle with a monocular camera is used to collect image information for detecting and counting the vehicles in different lanes and estimating their distance and speed on the defined route. The results indicate that our proposed image processing algorithms can acquire the information needed for intelligent traffic management systems.

Geolocation estimation of target vehicles using image processing and geometric computation

“…We trained YOLO-v3 on the KITTI dataset [25], as this study focuses on traffic objects, and KITTI includes eight categories of traffic objects: car, van, truck, pedestrian, person sitting, cyclist, tram, and misc [25]. To estimate the distance d from the ego vehicle to the target vehicle, we followed the approach proposed by Namazi et al [8].…”

“…This approach [8] was based on the pinhole camera model by considering the real, pre-known size of the target vehicle and the size of the bounding box added by YOLO-v3 around the target vehicle on the image plane. Distance d was calculated based on the average of the computed distances for both vehicle width and vehicle height by using a weight factor (i.e., 85% of the height and 15% of the width).…”

“…To detect lanes, as we presented in [8] and [9], we used canny edge detection [27] and the progressive probabilistic Hough transform [28][29].…”

“…In our previous studies [8][9], we have investigated the feasibility of using a vehicle equipped with a low-cost front-facing monocular camera with a built-in global positioning system (GPS) receiver (hereafter, we call this vehicle an ego vehicle) to observe another vehicle (hereafter, we call this vehicle an target vehicle) and estimate its speed, distance, and the lane it is in. After studies [8][9], a follow-up research question has been raised about the use of an ego vehicle in estimating the geolocation of the target vehicle, as accessing the vehicle's geolocation plays a critical role in modeling the traffic scene and making smart decisions by ITMSs.…”

A Systematic Review of Existing Methods plus the State-of-the-art Technologies Used to Collect Road-Related Data and its Usage

“…Given that the shadow is likely to be recognised as a vehicle due to the reflection and the different lighting situations, the system considers the shadow effect on data accuracy by background subtraction algorithm based on computing techniques to achieve efficient traffic management. Camera mounted on vehicle was introduced by Namazi et al [12], who proposed a system to detect the distance between vehicles and the speed of vehicles from multiple lanes. This dynamic detection achieves a large-scale inspection, which can detect the actual state of vehicles and the traffic condition on a real-time basis.…”