Ethiopian License Plate Detection and Recognition Using Deep Learning

Abstract

The study on License Plate Recognition (LPR) systems for Ethiopian to efficiently detecting and recognizing license plates from images. Our study aims to investigate and improve the performance of LPR systems by addressing key challenges, such as variations in license plate appearance, occlusions, and low image quality. The study uses a deep learning approach to design and develop an efficient LPR system. Three steps make up a typical LPR system: segmenting characters, detecting license plates, and recognizing characters. In our method, we localized the license plates while reducing the false positives by licensing detection. This study looks into the automatic localization of license plates using the state-of-the-art Mask R CNN object detector. Then, we Apply preprocessing techniques such as resizing, normalization, and image enhancement (e.g., contrast adjustment, noise removal) to enhance the quality of the license plate region. Finally, we feed the segmented character plate number to the CNN based character recognition model. In the study we collected 1190 plate images were fed into the system to assess the method's effectiveness, and the approach produced a 96. % accuracy rate for license plate recognition. An accuracy of 88 %was attained when all the recognized license plates were assessed to see how well the segmentation approach worked.