Nofima and SINTEF, in cooperation with QVision, have investigated the possibility of measuring the fat content of inhomogeneous pork trimmings in large plastic boxes for a Spanish meat company, Frigorificos Andaluces De Conservas De Carnes SA (FACCSA). These boxes are valued based on fat content; the lower the fat content, the higher the purchase value. Customers have different demands in terms of fat content and therefore knowledge of the leanness of each box is advantageous to both supplier and customer.
Current techniques for measuring fat content of meat trimmings rely on manual sampling or grinding of the meat. Manual sampling involves taking a sample from the box and cutting away all of the fat. The fat and remaining meat are then weighed separately and the percentage of fat is calculated. This is done sporadically throughout the process as is time-consuming and destructive. Grinding and blending the meat allows NIR to be used in reflection mode as the meat is more homogenous but this is not suitable to applications where the meat is sold as trimmings. The main problem with using NIR reflection measurements on inhomogeneous samples is that it is more improbable that the surface is going to be fully representative of the rest of the sample underneath. However, with transflection we probe deeper into the meat and fat and more absorption occurs thus giving more chemical information.
Physical scattering of the light is also a problem in this application as the consistency of fat and meat differ but more so because it is unavoidable that height variations will occur to some degree within the box. If the surface of the box has excessive peaks and valleys more physical scattering errors are introduced due to variations in optical path. Extended Multiplicative Signal Correction (EMSC) was successfully applied to the spectral data as a form of scatter correction and a Partial Least Squares model was built with an RMSEP (root mean square error of the prediction) of 1.9.
This model has been validated online with meat of different fat levels, various heights and in different colour boxes (red, blue, brown, white and yellow).
Current investigations are now underway in applying the same technology to classifying beef trimmings for Nortura BA. This application differs slightly in that the meat will be classified and sorted before it goes into boxes to gain more control over the sorting process. At present the meat is sorted manually by cutters who are trained to visually identify the fat content but this method is very prone to human error.
