License Plate Localization and Character Recognition using YOLOv8 and Convolutional Neural Network

The important role of the transportation system has an impact on increasing its use so that it has implications for security and traffic order issues. On the other hand, using technology such as a system for reading and recording vehicle license plates positively impacts this problem. From several research results related to digital image processing, YOLOv8 performs well in object detection speed and accuracy, while CNN can classify highly accurate images. In this research, YOLOv8 architecture is used to license plate localization, contour detection to segment license plate characters, and the CNN algorithm to classify segmented characters. In addition, dataset collection and image augmentation were carried out before training the model. Finally, it tested for the entire model using a pipeline with several dataset conditions and testing aspects. From the test results, the F1 score on the YOLOv8 model was 0.85 with a speed of 46 FPS, character segmentation with an F1 score of 0.48, and the CNN model obtained an accuracy of 60%.