Early Detection and Tracking of Distant Incoming Traffic using Improved Detection on Road Vanishing Point Reference for Adaptive Traffic Light Signaling

Abstract

Real-time monitoring is essential and influences the decision-making process of adaptive traffic light

systems. During temporary road closures, only one side of the lane can be accessed, increasing the need to

recognize and track oncoming vehicles. Therefore, it is crucial to detect oncoming vehicles that are far away as

early as possible, as waiting for an oncoming vehicle near a traffic light may delay the signal, leading to sudden

braking or an accident. The purpose of this study was to improve traffic detection and tracking, even when the

traffic is still far from the traffic lights. Vanishing point as detection reference is estimated, and Region of

Interest (RoI) is calculated. An evaluation is performed based on how quickly the proposed method detects

oncoming traffic compared to the R-CNN method. The results show that the proposed method requires an

average of 17.75 frames to detect the target vehicle, while R-CNN requires an average of 63.36 frames to detect

the target vehicle. The results show that the accuracy of the proposed method depends on the number of pixel

orientations when estimating the vanishing point and how accurately the RoI is defined. Therefore, the proposed

method reliably supports the safety and reliability of adaptive traffic light systems.