This study was carried out to investigate the potential of NIR
hyperspectral imaging system for non-destructive classification of
different pork grades. Experimental results have shown that the
traditional method used for pork classification can present some
weaknesses. These methods rely on average measurements of smaller
regions of the pork sample and are not suitable for assessment of every
sample in fast-paced large-scale production. Also, samples exhibiting
different classeswithin the loin eye region can be inaccurately classified
according to the location of the measurements acquired. The results
emphasized that the NIR hyperspectral imaging in the 900–1700 nm
range has the potential to classify pork sampleswithout any background
of physicochemical information. Multivariate techniques were used to
reduce the spectral dimension of the hyperspectral image data and
extract useful image features that are valuable for differentiating pork
grades. Distinctive spectral difference among pork grades could be
explained in various wavelengths in NIR range of the spectra. These
identified wavelengths are related to water and other chemical
components of the samples. This procedure allows for the later
introduction of cheaper multispectral NIR instruments for the desired
application. This study illustrated more accurate determination of pork
qualities by using non-destructive and chemical-free methods. For
industrial applications, it is necessary to implement on-line prediction of
intact meat to allow rapid and accurate quality assessment. The major
barriers are the high dimensionality of the hyperspectral data aswell as
the correct selection of representative ROIs. The first confronted
constraint could be overcome by the ideal selection of the most
important wavelengths; meanwhile the selection of ROI could be
performed by applying several image processing regimes on the
selected wavebands. Further studies should be carried out to in this
sense to overcome these disadvantages for proper industrial imple-
mentations. Generally, results highlighted from the recent work have
shown the potential of this technology to fulfill the need of the pork
industry for an accurate and fast method for quality classification
although some modifications must be supplied.