Applying data mining and Computer Vision Techniques to MRI to estimate quality traits in Iberian hams

Abstract

This study aims to forecast quality characteristics of Iberian hams by using non-destructive methods of analysis and data mining. Magnetic Resonance Imaging and Computer Vision Techniques were conducted on hams throughout their processing. Physico-chemical parameters were also measured in these products. Information from these analyses was integrated in a database. First, deductive techniques of data mining were applied to these data. Multiple linear regression allows for the estimation of information from Magnetic Resonance Imaging, Computer Vision Techniques and physico-chemical analysis. This enables the completion of the initial database. Then, predictive techniques of data mining were applied. Both, multiple linear regression and isotonic regression achieved the prediction of weight, moisture and lipid content of hams as a function of features obtained by Magnetic Resonance Imaging and Computer Vision Techniques. Thus, data mining, Magnetic Resonance Imaging and Computer Vision Techniques could be used to estimate the quality traits of Iberian hams. This allows for the improvement of the process control without destroying any piece.