A Unified AI-Based Vision System for Dehazing, Number Plate Recognition, and Traffic Signal Automation

Abstract

A Unified AI-Based Vision System for Dehazing, Number Plate Recognition, and Traffic Signal Automation" tackles hazy traffic conditions with cutting-edge AI, ensuring clear visibility and real-time traffic analysis through advanced dehazing and object detection methods, including Automatic Number Plate Recognition (ANPR). This allows the system to identify and track various objects on the road, including vehicles, pedestrians, and traffic signals. It surpasses traditional approaches by offering adaptive solutions to traffic congestion and leverages YOLOv5 for robustness in diverse environments. To overcome limitations in training data, a common challenge in intelligent transportation systems, it implements transfer learning. By fine-tuning a pre- trained YOLOv5 model on extensive, manually annotated datasets encompassing various traffic scenarios, the system achieves superior detection accuracy and execution time compared to traditional methods. This makes real-time analysis and congestion reduction possible, opening the door for safer, more effective transportation networks. Additionally, the system will incorporate an automatic number plate recognition (ANPR) module, enabling vehicle identification for various applications such as toll collection, traffic enforcement, and stolen vehicle detection.