Enhancing hard surface cleaning analysis using an open-source machine vision technique

Abstract: Evaluating the effectiveness of hard surface cleaning without bias is a challenging task. Traditionally, testing includes counting pixels in images taken during the surface cleaning process or assessing radiolabeled soils after removal. Of these two methods, pixel counting is effective and simpler; however, it only provides a binary clean/dirty result for each pixel.
 
The pixel counting method is enhanced with the incorporation of machine vision, which standardizes the color measurement relative to a baseline using Python libraries such as OpenCV, to analyze a commercially available standard IKW kitchen soil on a melamine tile. This new approach refines the “S” curve shape of the pixel counting method and yields a more precise intensity-based relationship between cleaning efficiency and surfactant. This approach is not only more accurate, but significantly reduces the effort and time required for analysis as the analysis can be completed 3-5 times faster than the pixel counting method and does not require the cumbersome equipment required for radio label analysis.