Real-time screening of melamine in coffee capsules using infrared thermography and deep learning

Abstract

Food adulteration is a major concern in the food industry, particularly in widely consumed products such as coffee. This study presents a novel non-destructive approach for detecting melamine contamination in coffee capsules using infrared thermography (IRT) and convolutional neural networks (CNNs). Coffee samples (natural, blended, and decaffeinated) with different coffee-to-milk ratios (1:3, 1:1, and 3:1) were adulterated with melamine at 2.5, 5, and 7.5 ppm. A dataset of 24,296 thermographic images was analyzed using ResNet34, achieving a classification accuracy of 95.71 % in blind validation. Compared to conventional chemical methods, this approach is faster, cost-effective, and scalable, making it a valuable tool for real-time food safety screening. The proposed method offers a non-invasive and rapid alternative to conventional analytical techniques such as HighPerformance Liquid Chromatography (HPLC) and Mass Spectrometry (MS), making it highly suitable for realtime quality control in the food industry