Prediction of pork quality characteristics using visible and near-infrared spectroscopy

Visible/near–infrared (Vis/NIR) spectroscopy was tested as a non–destructive method to assess the quality of fresh pork loin. Quality characteristics of pork were measured to develop calibration models for a diode–array Vis/NIR reflectance spectrophotometer. Spectral measurements and measurements for pH, water–holding capacity (WHC), XYZ tristimulus values, cook yield, shear force, proximate analysis (moisture, fat, and protein content), and subjective color, marbling, and firmness scores were performed for 411 fresh pork loin samples. Approximately 80% of the sample set was used for calibration and 20% for validation. Partial least squares (PLS) calibration models were developed for each characteristic measured. XYZ tristimulus values were predicted well with R2 > 0.91 and SECV 0.88 and SEP < 1.2 for validation. Moisture, fat, and protein content were predicted less well with R2 > 0.69 and SECV < 0.007 for calibration, and R2 > 0.61 and SEP < 0.007 for validation. The calibrations for the other characteristics (shear force, cook yield, pH, water–holding capacity, and scores for color, marbling, and firmness) varied from marginally good to poor with R2 values ranging from 0.584 to 0.002.