Beef quality evaluation using near infrared hyperspectral imaging

Abstract

To evaluate meat quality is always great concern at all stages of the food industry. Increasingly, industry requires methods for real-time quality control of fresh meat, in order to improve production efficiency, to ensure homogeneity of the products and to meet consumer expectations. In this paper, the imaging spectroscopy techniques, a tool that combines both, spectroscopy and computer vision advantages, will be used to improve the efficiency of beef quality evaluation. This research project aims to use the near infrared hyperspectral imaging to predict tenderness, chemical composition and sensory attributes of Nellore meat . It is also objective of this study to build distribution maps of these characteristics so we can study how the variability of each one of them occur between a sample and another and even within the same sample. Therefore, 94 Longissimus dorsi (sirloin) and 94 Biceps femoris (Picanha) of Nellore cattle from feedlot finishing system were used. Each sample was assessed following traditional methodology (reference analysis) for tenderness, dry matter, crude protein, fat and ash. pH and temperature, 1 and 24 hours post mortem were also measured. For each sample, a hyperspectral image was acquired in the spectral range 1000-2500 nm. The spectral and spatial data will be analyzed by chemometric multivariate data analysis and predictive models will be built for each of the characteristics in study. Is expected the models generated in this study to be effective to classify qualitatively and quantitatively beef samples and yet that the models to be used for chemical and tenderness imaging construction, which will possibly enhance the understanding of the characteristics variability within and between samples. (AU)