MADS

Problem

Urea is one of the most common and dangerous milk adulterants in India — 68.4% of 1,791 milk samples tested across 33 states were affected — and standard field tests only give a yes/no answer. Quantifying the concentration cheaply and quickly is the hard part.

Approach

Evaporate a few drops of milk on a heated plate (100.5 °C, 10–15 seconds), illuminate the residue with UV LEDs so the urea particles stand apart from other milk solids by refractive index, and film it through a 1000X USB microscope. A Python + OpenCV pipeline thresholds the footage in HSV space, contours the urea particles, and averages the measured concentration across 720 video frames.

What I built

• Complete hardware build on an Arduino Nano — PTC heating element, temperature sensor, UV LEDs, and a 1000X-zoom USB microscope — for ~₹3,140 in parts.
• A 4–5-drop milk sample is heated to 100.5 °C, evaporating the water content in 10–15 seconds; under UV light the remaining urea particles stand apart from milk proteins and carbohydrates by their differing refractive indices.
• Proprietary Python + OpenCV/NumPy pipeline isolates the illuminated urea with HSV thresholding, draws contours around the particles, and computes their area to quantify concentration.
• MADS v2 scans a 30-second video (720 frames at 24 fps) and averages concentration across frames — cutting the average error rate from 1.47% (v1, single static image) to 0.018%.
• Peer-reviewed publication in IJISRT (Vol 9, Issue 8; September 2024); pitched live at Nestlé's Moga factory to QA leadership, earning a commendation letter from Nestlé Corporate Affairs.

Result

Quantitative urea detection with a 0.018% average error rate on a ₹3,140 device — a peer-reviewed IJISRT publication, a $1,500 NFTE World Series of Innovation win, 1st place in the MetLife Foundation Good Health & Wellbeing Challenge, and a live pitch at Nestlé's Moga factory that earned a corporate commendation letter.

Media

Links