Medicine Tablet Authentication Using 'Fingerprints' of Ink-Jet Printed Characters

摘要

Counterfeit drugs have been a serious problem causing damage to people's health over the world. Numerous anti-counterfeiting methods based on tagging have been proposed; however, they suffer from three major issues: (1) tagging is applicable only to packages, not tablets directly; (2) end-users, i.e., patients, cannot inspect the tags; (3) tagging incurs extra costs for manufacturers. This paper describes a new method that we propose for authenticating individual medicine tablets as-is by matching images of printed characters. The printed characters on individual tablets of the same medicine seem the same to human eyes, but each is characterized by tiny unique differences. The contributions of this paper are: (a) to reveal the uniqueness of the characters printed by an actual pharmaceutical-use machine and (b) to propose a practical system to identify individual tablets using image matching. Our system is useful for any patients who want to authenticate a medicine tablet at hand: it only requires a picture with a smartphone camera. Our system is also useful for medicine manufacturers, because the database can be constructed using the existing manufacturing process without incurring additional cost. Our image matching algorithm recognizes very detailed features of the images and is accurate and fast even for a large-scale database. In conducted experiments, 1,000 sample tablets were captured using the same optical setup as an actual medicine manufacturing machine. Obtained results showed that 100% accuracy in individual tablet authentication was achieved.