A global calibration model for prediction of intramuscular fat and pH in red meat using hyperspectral imaging

Abstract

This study demonstrates a novel approach to develop global calibration models for predicting intramuscular fat (IMF) and pH across various red meat species and muscle types. A total of 8 hyperspectral imaging (HSI) datasets were used from different experiments, comprising data from three species: beef, lamb and venison across various muscle type, slaughter season and measurement conditions. Prediction models were developed using Partial Least Squares Regression (PLSR) and Deep Convolutional Neural Networks (DCNN) using a total of 1080 and 1116 samples for IMF and pH, respectively. Models for pH and IMF via both techniques yielded high Rc2 (0.86–0.93) and low SEC values. Also, reasonably accurate prediction performance was observed with high Rp2 (0.86–0.89) and low SEP values. Overall results illustrated the comprehensiveness of these global calibration models with the ability to predict IMF and pH of red meat samples irrespective of species and muscle type.