< Back to previous page


Prediction of pork meat quality parameters with existing hyperspectral devices

Book Contribution - Chapter

Prediction of pork meat quality parameters with existing hyperspectral devices B.H.R. Callens1, S. De Smet2, S.R. Cool1, F. Castaldi1, R. Van De Vijver1, E. Kowalski1,2 and M. Aluwé1 1ILVO (Flanders Research Institute for Agriculture, Fisheries and Food), Technology and Food Science Unit, Burgemeester van Ghansberghelaan 115, 9820 Merelbeke, Belgium, 2Ghent University, Department of Animal Sciences and Aquatic Ecology, Coupure Links 653, 9000 Ghent, Belgium; bert.callens@ilvo.vlaanderen.be In this study, the use of two commercially available hyperspectral devices for measuring fresh pork quality traits was investigated as a faster measurement for different fresh meat quality reference parameters. One VNIR linescan hyperspectral camera (CAM) Specim FX10e™, 400-1000 nm and one hyperspectral spectrometer with a probe optimized for meat measurements (SPEC) ASD Labspec 4™ with Bonzai Advanced Meat ProbeTM, 350-2,500 nm, were tested. On 5 days, each time 6 batches of carcasses with different genetic background were selected in the slaughter house. Per batch, four carcasses (two male, two female) were chosen with high and low lean meat percentages, resulting in 120 animals in total of which the Longissimums thoracis et lumborum (LTL) was collected and cut 24 hours after slaughter. Spectra from the region of interest on a transverse cut surface were obtained from the hypercubes of the VNIR camera with image processing for three slices per LTL. The spectra were linked to the reference data (colour, pH, shear force, drip loss, IMF, sensory panel) with partial least squares regression. Based on Ratio of Performance to Deviation (RPD) for the prediction (Pred), the models showed promising performances for CIELAB colour for CAM, with (RPD Pred L*=2.19, a*=2.08, b*=1.72) and SPEC (RPD Pred L*=1.60, a*=1.61, b*=1.76) at individual sample level. For CAM, ultimate pH and intramuscular fat models also had RPD promising models with RPD Pred 2.13 and 1.82 respectively. For SPEC, it was necessary to average the spectra and reference measurements on animal level (RPD 1.84 and 1.65) in order to have good predictions. Drip loss and shear force showed clear correlations, but were too unstable to have good predictions. Models for cooking loss and sensory panels results showed low correlations (RPD Pred <1.3). Overall, hyperspectral measurements show potential as in- or at-line (sample based) implementation, although differences between the sensors were found and optimization remains necessary.
Book: Book of Abstracts of the 72nd Annual Meeting of the European Federation of Animal Science
Pages: 541
Number of pages: 1
ISBN: 978-90-8686-366-2
Publication year:2021