Automated Method for Determination of Cheese Meltability by Computer Vision

Meltability is the property of a cheese to flow and spread, as well as the loss of the integrity of the cheese structure by heating, and is of great importance for cheeses used as ingredients in other products. Analytical methods for determination of cheese meltability are laborious and time-consuming, requiring heating of cheese samples followed by assessment of the dimensional changes, such as diameter and height. In this study, computer vision is proposed for rapid and accurate determination of meltability for five different types of cheese and compared to the reference Schreiber method. Digital images were acquired from the samples before and after melting, and the variation of samples was measured after segmentation and identification of the region of interest (ROI). The meltability measured by the Schreiber method varied from 0 to 45.37%, while the results determined by the computer vision method varied from 0.37 to 99.64%. The computer vision method presented high correlation with the Schreiber method, providing better results for samples that had irregular shape after melting, when compared to the traditional method. Additionally, the correlation between the manual and the automatic methods was calculated and the results showed a perfect correlation between all the computer vision methods (> 0.99). The results indicate the potential application of computer vision as a standard analytical method for determination of cheese meltability.