Machine Learning and AI-Driven Smart Sensing Technologies for Accurate Detection of Adulterated Food Products

Authors:

Abstract
This study suggests an AI-based multimodal smart sensing system to accurately and timely detect adulterated food products using intelligent sensing and machine learning technologies. The framework is a fusion of the four techniques of spectroscopy, hyperspectral imaging, impedance sensing, gas sensing, and the 5th of IoT-based environmental monitoring within an adaptive multimodal learning architecture. The experimental data were obtained by spiking milk, honey, turmeric powder, coffee powder, olive oil, red pepper powder and processed meat products with varying environmental conditions. For classification, deep learning, ensemble learning, and adaptive feature fusion techniques were used and for real-time deployment on the edge, deep learning was used. With the proposed framework, classification accuracy was 98.14% on average, and the robustness was excellent in noisy and dynamic operational conditions.

This publication has 30 references indexed in Scilit: