Hyperspectral Identification of Milk Adulteration Using Advance Deep Learning
Muhammad Aqeel,Ahmed Sohaib,Muhammad Iqbal,Syed Sajid Ullah
DOI: https://doi.org/10.1109/access.2024.3504334
IF: 3.9
2024-12-04
IEEE Access
Abstract:Food adulteration poses significant health risks globally and is rigorously monitored by safety authorities. In developing nations, where milk is highly prone to contamination (with Brazil, India, China, and Pakistan producing half of the world's milk), stringent detection and classification techniques are essential. This study employs both destructive and non-destructive methods for milk adulteration analysis. The destructive method uses Lactoscan for comprehensive qualitative measurements, including temperature, pH, conductivity, solids, protein, density, fat content, and SNF. The non-destructive method utilizes hyperspectral imaging (HSI) with the Specim Fx-10 (397–1003 nm) for image-based analysis, involving preprocessing steps like image scaling, ROI selection, radiometric correction, and spectral reflectance extraction using the empirical line method (ELM). Advanced deep learning models, including Convolutional Neural Networks (CNN), Artificial Neural Networks (ANN), Long Short-Term Memory networks (LSTM), and Gated Recurrent Units (GRU), are employed to predict and classify pure and adulterated milk spectra. CNNs showed superior performance in identifying adulteration trends. The proposed pipeline, validated with a 97% accuracy, outperforms state-of-the-art techniques based on metrics such as Kappa, accuracy, precision, recall, F1-score, MCC, and Jaccard Index.
computer science, information systems,telecommunications,engineering, electrical & electronic