Urban traffic congestion continues to intensify due to rapid population growth, increased vehicle ownership, and inefficient fixed-time signal systems. Conventional traffic lights operate on predetermined cycles that fail to adapt to real-time traffic variations, resulting in longer waiting times, fuel wastage, and increased environmental pollution. This research proposes an embedded deep vision module for dynamic traffic signal control based on real-time density recognition. The system employs deep learning–based object detection models integrated with embedded hardware to identify vehicle count, classify traffic density levels, and adjust signal timing automatically. The module processes live video feeds from traffic cameras and uses convolutional neural networks (CNNs) to detect and quantify vehicles accurately under varying lighting and weather conditions. The processed data is transmitted to a microcontroller-controlled signal unit, enabling optimized signal intervals that reduce congestion and improve traffic flow efficiency. The proposed system offers scalability, low latency, and cost-effectiveness suitable for smart city deployments. By leveraging deep vision and embedded intelligence, the approach enhances urban mobility, reduces emissions, and provides a robust infrastructure for future intelligent transportation systems.
IJESAT (Tue,) studied this question.