In this paper (open access) the authors present “PowDew”, an early-stage commercial system designed to detect counterfeit powdered infant formulas using only a commodity smartphone camera.
It works on the principle that different powdered formulas exhibit unique properties upon contact with liquid, discernible through a water droplet motion interacting with the powder., PowDew analyzes the droplet’s spreading and penetration, to infer information correlated to the powder properties such as wettability and porosity, which are key indicators of the formula’s authenticity.
The authors conducted real-world experiments under varying conditions with different brands of powdered infant formula and adulterants. They reference a resultant 12,000 minutes of video recordings of the droplet motions on various infant formulas, including authentic and altered. The programme uses machine learning to extract features from the video frames.
They report that PowDew yielded an overall detection accuracy of up to 96.1% for this application, and consider that it could be trained on models for other applications by either industry QC testing or regulatory authorities..
Photo by Daniel Romero on Unsplash
Comments