<![CDATA[Identity and vehicle ownership checks are an important routine carried out by security personnel at the State Islamic University of North Sumatra (UINSU) to ensure the legality of vehicles entering and exiting. However, the still manual inspection process of matching vehicle registration certificates (STNK) and license plates is often time-consuming and causes vehicle queues, especially during peak hours. This research applies a combination of the Canny Edge Detection algorithm to detect license plate edges and Tesseract OCR (Optical Character Recognition) to extract text from images. The purpose of this research is to explore the effectiveness of this method in detecting and recognizing two-wheeled vehicle license plates within the UINSU environment, and to provide an alternative solution to the problems of the manual inspection system. The dataset used consists of 125 images of two-wheeled vehicles taken using a mobile phone camera. The research results show that before post-processing was applied, OCR produced an average character accuracy of 81.60% with a CER of 18.40%, while after post-processing, the accuracy increased to 82.49% and the CER decreased to 17.51%. These results confirm that rule-based correction is able to improve character reading errors, although the improvement is moderate and has not completely addressed cases of detection failure in some images. This finding serves as the basis for addressing queuing issues and human resource limitations, while also providing a foundation for the broader development of digital image-based vehicle identification systems across various sectors.]]>
Copyrights © 2025
