<![CDATA[The exchange of personal data in Indonesia remains predominantly manual, involving form-filling and photocopying of electronic identity cards (e-KTP), despite the availability of embedded electronic chips designed for automated data processing. This study proposes an integrated data extraction and validation system combining YOLOv8 for precise region detection and Optical Character Recognition (OCR) with advanced preprocessing techniques for textual information extraction. Unlike previous approaches relying solely on OCR (e.g., Vision AI), this method employs YOLOv8 object detection to accurately localize key fields (NIK, Name, Address) before text extraction, followed by validation through the DUKCAPIL API. The system was evaluated using 20 e-KTP images captured under various conditions. Results demonstrate that the proposed approach achieves an average OCR accuracy of 98.7% with an Intersection over Union (IoU) of 0.975, significantly outperforming baseline Vision AI extraction by 15–20%. All extracted data successfully passed validation against the official DUKCAPIL database, confirming 100% authenticity verification. This system provides an economical and efficient solution for automating population data administration, particularly suitable for small non-governmental organizations with limited budgets. The integration of deep learning-based object detection and preprocessed OCR offers a robust framework for digital identity verification systems.]]>
Copyrights © 2025
