Medicine Authentication Based on Image Processing Using Convolutional Neural Networks

The proliferation of counterfeit pharmaceuticals presents a critical challenge to public health and safety worldwide. A common method to differentiate authentic from counterfeit medicine is visually inspect the medicine's packaging. However, medicine packaging is susceptible and easy to replication. This research investigates an image-based classification method using machine learning and computer vision to differentiate between counterfeit and authentic medicine packaging. This research aims to develop a system that can authenticate local paracetamol and Biogesic based on support vector machines and convolutional neural networks. This can be done through these specific objectives: (1) To be able to capture PNG images of the packaging of the paracetamol using a Raspberry Pi camera; (2) To be able to use Support Vector Machines for feature extraction and Convolutional Neural Network models for the classification; (3) To be able to verify the reliability of the system using a confusion matrix. The classification of this study was evaluated using a confusion matrix. The resulting accuracy of the system is 88.75 %. Despite the promising results yielded by the developed system, the researchers recommend considering a wider database of medicines as well as features to consider.