<![CDATA[This research develops an Optical Character Recognition (OCR) application to recognize and translate Arabic-Malay script, widely used in Bengkalis and surrounding regions. The application is designed to facilitate learning and interaction with the script through features for text recognition and translation. To enhance OCR accuracy, advanced image preprocessing techniques and a comprehensive training dataset were employed. Testing under various lighting conditions showed an average recognition accuracy of 25%. Performance was optimal under well-lit conditions, achieving 100% accuracy at 1639 lux, but accuracy significantly declined in low-light environments, particularly at 3.5 lux, where it dropped to 0%. While this result demonstrated the system’s potential, a comparison with similar OCR systems indicates significant room for improvement, particularly in recognizing complex or handwritten Arabic-Malay characters. The results highlight challenges related to lighting conditions, including variations in light sources and angles, which greatly affect recognition accuracy. Indeed, future work should focus on expanding the training dataset with diverse samples, refining preprocessing techniques such as automatic brightness adjustment, and implementing advanced machine learning models like Convolutional Neural Networks (CNNs). This application holds significant potential for practical uses in education, translation services, and local communication, contributing to preserving and promoting the Arabic-Malay script as part of cultural heritage in the modern era.]]>
Copyrights © 2025
