Infrared thermography of beef carcasses and random forest algorithm to predict temperature and pH of Longissimus thoracis on carcasses

Abstract

This study aimed to evaluate the use of infrared thermography (IRT) as a method for predicting the initial and ultimate temperature, as well as the pH, of the Longissimus thoracis in beef carcasses (LTBC). A total of 102 beef carcasses, consisting of 62 F1 Red Angus × Nelore and 40 Nelore cattle, approximately 14 months of age, were evaluated before and after refrigeration. Temperature and pH values of the LTBC were measured using a probe thermometer and a pH meter. To predict these parameters, IRT was used to measure surface temperature features of the fore, hind, and Longissimus thoracis regions, as well as the whole carcass. The Random Forest machine learning algorithm was applied for predictive modeling. The results of this study indicated that it was possible to use IRT to predict temperature of LTBC with R² values ranging from 0.06 to 0.78 and MAE from 1.12 to 1.67. For initial temperature was R² of 0.06 and ultimate temperature with R² values 0.26 and 0.17. The inclusion of hot carcass weight (HCW) parameter improved the prediction of ultimate temperature with an R² from 0.78 to 0.86. The prediction of LTBC pH with thermal images showed R² values ranging from 0.26 to 0.82 and MAE from 0.10 to 0.82. The combination of IRT and HCW improved the prediction of muscle ultimate pH in the carcass. In conclusion, the IRT method can predict the initial and ultimate pH and temperature of LTBC, with improved accuracy when combined with the HCW parameter.