Smart Traffic Light System Using YOLO-Based Vehicle Detection for Real-Time Traffic Density Analysis

Abstract

Traffic congestion at signalized urban intersections remains a major challenge due to dynamic vehicle volume and limited road capacity. Fixed-time traffic signals often fail to respond to real-time traffic variation, causing inefficient signal distribution and longer vehicle queues. This paper proposes a YOLOv10-based traffic monitoring framework combined with a Multi-Context Adaptive Signal Weighting (MC-ASW) method for adaptive traffic signal simulation. Vehicle count obtained from video-based car detection is used to classify traffic density, while contextual variables such as time of day, weather condition, and road condition are incorporated into an adaptive score for signal timing decisions. Green-light duration is adjusted according to the resulting adaptive score, while yellow-light transition, all-red clearance, and phase-dependent red-light behavior are considered to maintain traffic signal safety. The proposed framework is evaluated through a simulation-based experimental design using detection performance metrics and traffic signal performance indicators. This study focuses on lightweight car detection and adaptive signal timing optimization as the main research contributions.