Density Based Traffic Signal System Using AI

Abstract

Traffic congestion in urban areas is largely caused by traditional traffic signal systems that operate on fixed timing mechanisms without considering real-time traffic conditions. This project proposes an AI-based adaptive traffic signal control system that dynamically adjusts signal timings based on real-time vehicle density using computer vision and deep learning techniques The system utilizes a Hybrid YOLOv8 model with multi-threading to detect, classify, and count vehicles from live traffic camera feeds. Lane-wise vehicle density and average waiting time are computed and processed by an intelligent control algorithm to allocate green signal durations dynamically. An emergency vehicle detection mechanism is integrated to automatically override normal signal operation, ensuring faster and safer passage for priority vehicles. A hardware prototype using a Raspberry Pi 5 and LED-based traffic signal representation validates the system's real-time adaptive behavior. Performance evaluation shows reduced vehicle waiting time, improved traffic flow efficiency, lower fuel consumption, and enhanced emergency response. The proposed solution is scalable, cost-effective, and suitable for integration into IoT-based smart city traffic management systems.