Real-Time Detection of Milk Adulteration with a Portable Multispectral Analysis Device: A Multispectral Sensor and Optimized Logistic Regression Approach

This study presents the development of a portable, low-cost, and edge computing-based system for real-time milk adulteration detection. Utilizing an AS7265x multispectral sensor and Arduino Nano 33 BLE Sense microcontroller, this system employs an optimized logistic regression model to identify starch adulteration in milk samples with near-perfect accuracy. Unlike complex neural network models, the logistic regression model offers simplicity, low power consumption, and efficient operation on microcontrollers. The collected spectral data is processed in real-time, and results are transmitted via Bluetooth for immediate analysis. The system demonstrates high accuracy, portability, and cost-effectiveness, making it suitable for use in various stages of the milk supply chain, including farms, processing facilities, and retail points. Future work will explore the detection of other adulterants and the integration of cloud-based analytics to enhance monitoring capabilities. This study provides an innovative approach to ensuring milk quality and consumer safety.