A dual-engine fusion optical character recognition method for fast identification and key information extraction of drug labels

Abstract

In the context of smart healthcare and information-driven drug supervision, the automatic recognition and extraction of drug label information presents a significant challenge. Traditional Optical Character Recognition (OCR) methods often struggle with complex backgrounds, diverse fonts, and mixed languages. This paper proposes a dual-engine fusion OCR method combining EasyOCR and CnOCR to enhance recognition accuracy. The method integrates IoT-based data collection for real-time drug information monitoring, utilizing multi-threaded parallel recognition for efficiency and an image preprocessing pipeline (including tilt correction, deblurring, and contrast enhancement). Additionally, a field area positioning and template matching mechanism ensures the precise extraction of key information such as drug name, ingredients, specifications, and expiration date. The approach achieves over 92% accuracy across various real-world scenarios, demonstrating improved robustness and promising potential for digital drug management, as well as IoT-based drug traceability and supervision.