The Milk Adulteration Detection System (MADS) is an innovative IoT and machine learning-based solution for real-time milk quality assessment, achieving 95% accuracy in detecting adulterants such as urea, starch, formaldehyde, sodium bicarbonate, and maltodextrin. Developed as a final year project, MADS addresses critical challenges in the dairy industry by providing a secure, scalable, and user-friendly platform for quality control.
Utilizes ESP32-interfaced pH, temperature, and gas sensors to capture live milk quality data, updated every 5 seconds on a Django web dashboard.
Employs a Random Forest Classifier trained on a Kaggle dataset (4243 samples, 9 features: pH, temperature, gas, fat, SNF, protein, EC, lactose, gravity) for precise adulteration detection.
Processes USB-based CSV data from the Lactoscanner S60 (fat, SNF, protein, EC, lactose, gravity) to overcome RS232 communication challenges, using Pandas for preprocessing.
Leverages SHAP (SHapley Additive exPlanations) to highlight key features (e.g., EC, pH, lactose) driving predictions, enhancing trust in results.
Features a Django dashboard with English and Urdu (Noto Nastaliq font) interfaces, ensuring accessibility for Pakistani dairy farmers.
Integrates Firebase Realtime Database for data storage and Firebase Authentication for role-based access (admin, quality control, user).
Displays live sensor trends and prediction outcomes using Chart.js, with Bootstrap 5 for a responsive, user-friendly design.
pH Sensor, Temperature Sensor, Gas Sensor, Lactoscanner S60, ESP32 Microcontroller, Arduino Uno, Breadboard.
Python 3.8+, Django 5.0, Firebase (Realtime Database, Authentication), Chart.js, Bootstrap 5, Jupyter Notebook.
scikit-learn (Random Forest Classifier, StandardScaler), Pandas, NumPy, SHAP, joblib.
AJAX for asynchronous updates, Noto Nastaliq Urdu Font, StandardScaler for data normalization.
Achieves 95% accuracy in detecting five common milk adulterants, validated on a balanced Kaggle dataset. Overcomes hardware limitations (e.g., RS232 garbage values) by implementing a USB-based CSV upload solution with Pandas preprocessing. Provides a scalable framework for dairy quality assurance, with potential for mobile app development and expanded adulterant detection (e.g., melamine, detergents). Supports Pakistan’s dairy industry with a multilingual interface and real-time insights, fostering consumer safety and industry trust.
