Comparative Evaluation of YOLO-based Models for Indonesian License Plate Detection and Recognition with EASY OCR

License Plate Recognition (LPR) plays a crucial role in intelligent transportation systems (ITS), enabling applications such as traffic law enforcement and smart city surveillance. This study compares the performance of YOLOv10n and YOLOv11n in detecting Indonesian vehicle license plates, integrated with EasyOCR for character recognition. The evaluation considers key metrics such as Precision, Recall, [email protected], and inference speed. YOLOv11n outperforms YOLOv10n with higher precision (0.9679 vs. 0.9615) and recall (0.9679 vs. 0.9259) during the validation process. On the test set, YOLOv11n achieved better average accuracy on original images (0.8508 vs. 0.8208) and processed images (0.8032 vs. 0.7919), with a higher F1-score (0.9679 vs. 0.9434). Additionally, fuzzy matching using Levenshtein distance for character recognition showed superior performance for YOLOv11n, with an average fuzzy matching score of 0.85 for original images and 0.80 for processed images, compared to YOLOv10n’s scores of 0.82 and 0.79, respectively. Both models exhibited a noticeable accuracy drop when using processed images, but YOLOv11n showed better resilience to preprocessing distortions. These results suggest that YOLOv11n is the better-performing model, offering superior detection, recognition, and character matching capabilities, especially for real-time LPR applications in resource-limited environments.