Integrating Cnns and Chemometrics for Analyzing Nir Spectra and Rgb Images in Turmeric Adulterant Detection Publisher



Sadeghi A¹ ; Khani S¹ ; Sabourian R² ; Hajimahmoodi M² ; Ghasemi JB¹ Show Affiliations
Source: Journal of Food Composition and Analysis Published:2025

Abstract

This study explores the detection and quantitative analysis of adulterants in turmeric using advanced optical sensing techniques, including Near-Infrared (NIR) Spectroscopy, RGB image analysis, and chemometric methods, coupled with deep learning models like Convolutional Neural Networks (CNNs). Turmeric, widely used in culinary and medicinal fields, is prone to adulteration with substances such as starches, flours, and synthetic dyes. Conventional detection methods are labor-intensive and impractical for large-scale screening. In this research, CNNs were applied to process large datasets, while NIR spectroscopy and RGB image analysis were integrated with chemometric models to enhance adulterant detection. A dataset was created using turmeric samples adulterated with corn starch, wheat flour, and rice flour at varying concentrations. Both Partial Least Squares Regression (PLSR) and CNN models were developed to predict adulterant concentrations, demonstrating strong correlations between predicted and actual values. A CNN-based classification model was also built to differentiate among nine concentrations of adulterants, representing various types and concentrations, with high sensitivity and specificity. The integration of optical sensing techniques and deep learning provides a non-destructive, rapid, and accurate method for detecting adulteration, offering significant potential for enhancing food safety and quality control in turmeric. © 2025 Elsevier Inc.