Enhanced License Plate Recognition For Stolen Vehicle Recovery: A Scalable and Cost-Effective Solution

This paper presents the development and implementation of a license plate recog-nition (LPR) system designed for integration with a commercial information system for stolen vehicle recovery. The study highlights the challenges of existing solutions, such as scalability limitations and high operational costs, particularly for vehicle tracking in high-demand scenarios. To address these issues, a collaboration between a Brazilian vehicle recovery company and the University of Pernambuco (UPE) led to the development of a new software solution. The proposed system leverages machine learning and computer vision techniques, incorporating OpenCV for video processing, YOLOv8 for vehicle and license plate detection, and EasyOCR for optical character recognition. The development followed an agile methodology, ensuring iterative improvements and real-world validation. Key challenges included handling low-resolution images, optimizing OCR accuracy, and ensuring system robustness under various environmental conditions. The study explores preprocessing techniques and algorithmic optimizations to enhance detection reliability. The results indicate that the new LPR system improves detection accuracy, reduces dependence on third-party infrastructure, and offers a scalable, cost-effective alternative for vehicle monitoring. Future work will focus on refining OCR accuracy, improving image preprocessing, and integrating advanced machine learning techniques for further performance enhancements.