Browsing by Author "Makhubo, Qhakazile Angel"
Now showing 1 - 1 of 1
Results Per Page
Sort Options
- ItemClassification of game meat species Springbok (Antidorcas marsupialis) and Blesbok (Damaliscus pygargus phillipsi) with NIR hyperspectral imaging(2020-12) Makhubo, Qhakazile Angel; Williams, Paul James; Manley, Marena; Hoffman, Louwrens C.; Stellenbosch University. Faculty of AgriSciences. Dept. of Food Science.ENGLISH ABSTRACT: Game meat is a lucrative alternative to domesticated meat. It is lower in fat and is an organic product, as it complies with organic agricultural enterprises. However, the game meat industry operates as a free-market in South Africa and is a high value product, making it a vulnerable target to food fraud. The development of methods to assess the authenticity of game meat would be a favourable investment for both the consumer and producers. Thus, this investigation aimed to classify different game meat species, namely Springbok (Antidorcas marsupialis) and Blesbok (Damaliscus pygargus phillipsi) using NIR hyperspectral imaging. Additionally, classification of the different muscles, namely the longissimus thoracis et lumborum (LTL), biceps femoris (BF), semimembranosus (SM), semitendinosus (ST), infraspinatus (IS) and supraspinatus (SS), and classification the species irrespective of the muscle was investigated. NIR hyperspectral imaging offers rapid measurement of data, and provide both spatial and spectral information. Two analytical approaches, namely pixel and object wise were investigated. Multivariate analysis techniques, namely principal component analysis (PCA) and partial least squares discriminant analysis (PLS-DA) were applied for exploratory and classification purposes, respectively. Object-wise analysis performed better in comparison to pixel-wise, with the classification accuracies of 95.90 and 71.49 % respectively. Additionally, the species could successfully be classified when using one muscle and when any muscle of the six muscles was used. Classification of the muscles was successful, however, grouping into hindquarter (containing the BF, ST, SM), forequarter (containing the IS and SS) and LTL had to be done to achieve this. Moreover, despite the successful classification, the other statistical parameters used to assess models were unsatisfactory.