Оптико-геометричні особливості упаковок лікарських засобів у задачах автоматизованого розпізнавання зображень

• Date: 2026
• Main Authors: Дацок, Є.О., Яковлева, О.В.
• Format: Article
• Language: Ukrainian
• Published: Vinnytsia National Technical University 2026
• Subjects: автоматизоване розпізнавання зображень; мультимодальні моделі; OCR; упаковки лікарських засобів; комп’ютерний зір; оптичні характеристики; обробка зображень; мультимодальний аналіз
• Online Access: https://oeipt.vntu.edu.ua/index.php/oeipt/article/view/867
• Journal Title: Optoelectronic Information-Power Technologies

The paper presents an analysis of the optical and geometric characteristics of pharmaceutical packaging in AI recognition tasks. The study considers the specifics of medication packaging as a complex object for automated image analysis, including the influence of geometric properties, reflective surfaces, small text, multilingual labeling, and illumination conditions on recognition quality. The limitations of classical OCR approaches for this type of packaging are analyzed, particularly those related to text deformation on curved surfaces, glare artifacts, low contrast, and complex image structures. Practical recommendations for photographing the packaging to improve recognition stability are also considered. The findings demonstrate that optical image characteristics significantly influence the effectiveness of AI-based analysis and should be taken into account during the design of multimodal recognition systems.