Improved identification and classification accuracy of wooden breast by jointly using near-infrared spectroscopy and compression speed

Abstract

A uniform standard for classifying wooden breast (WB) is still being explored. The palpation method is the most common grading method in WB, but it requires a large number of people, who need professional sensory training. This study aims to verify the feasibility of near-infrared spectroscopy (NIR) and compression speed for the classification of WB. The quality characteristics of different categories of WB and normal breast (NB) show that the hardness of raw WB is significantly higher than that of NB, but no difference was observed in the shear force between NB and WB. The water holding capacity (WHC) of NB is better, and the weight, height, redness (a*) and yellowness (b*) of WB are higher than those of NB. Afterward, the NIR and compression speed were used to identify and classify WB by using Partial Least Squares Discriminant Analysis (PLSDA). The results show that the NIR can effectively distinguish between NB and WB, the total classification accuracy of modeling and prediction are 85.71% and 81.58%, respectively. Then, for WB, the compression speed was further used to achieve precise classification and the total classification accuracy of modeling and prediction from PLSDA are 80.65% and 82.14%, respectively. To a certain extent, the compression speed inspired by ergonomics can quantify and simulate the palpation method. The results clarify that WB and NB can be identified by combining the compression speed and NIR, and it can be used as an assessment tool for food classification.

Keywords: Classification; Compression speed; Meat quality; Near-infrared spectroscopy; Wooden breast.