Hyperspectral Imaging—A New Era of Applications in Non-Destructive Sensing of Meat Quality

Abstract

Introduction T he high labour costs, inconsistency, subjectivity and variability associated with human evaluation of meat quality justify the need for objective measurement systems. A hyperspectral imaging system has been developed recently as a rapid and effective tool for meat quality assessment, allowing the determination of several quality parameters simultaneously. This technology combines the advantages of both spectroscopy and imaging techniques in one system to provide both spectral and spatial information for better and deep inspection of meat products. Coupled with robust multivariate analyses, the system has been used successfully for differentiation between meat cuts, distinguishing between different quality grades and discrimination between different species. The system has also been utilised for accurate prediction of water holding capacity, colour, pH, drip loss and tenderness, the most important quality traits in beef, pork and lamb. Moreover, the developed hyperspectral imaging system has been exploited not only for predicting major chemical constituents such as moisture, fat and protein in different meat species with reasonable accuracy but also for building chemical images to show the distribution maps of these constituents in a direct and easy way, which facilitates meat grading and labeling. The great potential of this technology elucidated from these research endeavours promotes this technology as a good candidate and a better choice for in-line or at-line assessment of meat quality non-destructively in the modern meat industry. The question that needs to be answered now is “how all of these tasks could be achieved using a clear-cut methodology”? The answer to this question is also straightforward “you need to have four things in your hands: (i) high-quality spectral images inside which the spectral fingerprints (X-matrix) of the meat samples being analysed reside, (ii) the reference values (Y-matrix) of all traits that you want to predict, (iii) a chemometrics tool (the corner stone) to incorporate this information in a multivariate structured model and (iv) an image processing tool to visualise the data in all pixels of the image to build ‘chemical images’ of the tested meat quality traits”. Let us see how one can effectively employ this strategy in a particular scenario.