فصلنامه تخصصی فناوری اطلاعات و ارتباطات
نوع مقاله : مقاله پژوهشی
نویسنده
خیابان ۱۹دی خیابان هلال احمر ۲۰متری حائری پلاک ۱۲۹
کلیدواژهها
عنوان مقاله English
نویسنده English
Background and Objectives: in this research, a novel method for detecting Iranian license plates and simultaneously recognizing letters and digits is introduced. Various approaches have been proposed for license plate detection, which have progressively improved over time. Traditional methods involve classical image processing techniques such as grayscale conversion, thresholding, and binary morphology. With advancements in machine learning, artificial neural network algorithms have been employed for license plate identification and recognition. Today, deep neural networks, particularly convolutional neural networks and recurrent neural networks, are recognized as leading methods in vehicle license plate detection. The objective of this study is to propose an efficient approach that includes locating the license plate using machine vision algorithms and segmenting the identifiers. Challenges such as variations in lighting conditions and vehicle speed have been examined in this research, and solutions to enhance detection accuracy and speed have been presented.
Methods: To improve the accuracy and robustness of the proposed model, a dataset consisting of 844 real Iranian license plate images captured under diverse lighting conditions and viewing angles was collected and expanded to 2024 images using data augmentation techniques. After fine-tuning, the model is capable of localizing license plate regions and extracting Persian characters using a CRNN-based architecture. The proposed system demonstrates a significant performance improvement over YOLOv8 and YOLOS models by reducing recognition errors in challenging scenarios such as strong illumination, shadows, high vehicle speed, and font variations
Findings: Ultimately, an accurate and efficient license plate recognition method based on the YOLOv9 architecture is proposed, achieving an overall accuracy of 98%. Quantitative evaluation results, including precision, recall, and F1-score, indicate that the proposed model is well suited for real-time applications in visual surveillance, traffic monitoring, and intelligent transportation systems.
Conclusion: The experimental results demonstrate that the proposed approach significantly enhances license plate detection accuracy and can effectively support intelligent transportation systems and automated traffic management applications.
کلیدواژهها English
