A three-staged approach to medicine box recognition

Abstract

This paper proposes a novel take on the task of medicine box recognition using a three-staged approach. We present the algorithms that will be used for each stage: barcode recognition, text recognition and feature matching techniques. The final goal of the system is to use a camera mounted on a device using an Android system to correctly identify medicine boxes and then provide people with difficulties (elderly, visually impaired) with relevant information about it. We report a good success rate of 80% in identifying boxes using their barcodes (in still pictures) or by running an OCR to detect the drug name. The success rate is higher (100%) in the case of feature matching, but the developed method is slow in its present form.