Automatic Vehicle License Plate Detection from Security Cameras using Deep Learning Techniques

License plate recognition is essential for identification, tracking, searching, and surveillance systems in Thailand's smart cities. This study addresses challenges such as the complexity of Thai characters and numbers on license plates and varying environmental conditions like lighting, angles, and occlusions. The research introduces a deep learning-based system to recognize Thai characters and numbers from high-security CCTV images. Utilizing models YOLOv5, YOLOv8, YOLOv9, and YOLOv10, the system detects license plates, segments plate components, and recognizes characters. The research demonstrates that YOLOv10 achieves a detection accuracy of 98.3%, image segmentation accuracy of 99.0%, and character recognition accuracy using Optical Character Recognition (OCR) exceeding 98.5%. Beyond these accuracy metrics, the unique contributions of this study include developing a robust ALPR system capable of handling the intricacies of Thai license plates and integrating advanced deep-learning techniques to enhance real-time processing and reliability. This system significantly improves the efficiency and reliability of license plate recognition, offering substantial benefits for traffic management, law enforcement, and security surveillance in smart city applications.