Evaluating YOLO11 and EasyOCR for Indonesian License Plate Recognition: A Study of Black and White Plate Designs

This study presents a comprehensive evaluation of an Automatic License Plate Recognition (ALPR) system tailored to the specific characteristics of Indonesian license plates, focusing on black and white background designs. Motivated by a regulatory shift in license plate color schemes, the research investigates how design differences impact recognition performance. The proposed system integrates YOLO11 for license plate detection and EasyOCR for character recognition, enhanced through an adaptive preprocessing pipeline that includes grayscale conversion, Gaussian blur, and Otsu thresholding with conditional color inversion. A dataset of 5,900 annotated images was collected from public sources and surveillance cameras under daylight conditions and distances ranging from one to two meters. Key challenges addressed include handling diverse lighting conditions, low-resolution plates, occlusions, skewed angles, and contrast variation between plate text and background. The proposed model achieved a detection precision of 88.3%, recall of 96.9%, and F1-score of 92.4%. Recognition accuracy improved significantly with preprocessing, reducing character error rates to as low as 3% on black plates. These results demonstrate the system's robustness and adaptability, highlighting the critical role of design-aware preprocessing and context-sensitive acquisition in improving ALPR performance in dynamic regulatory environments.