Meat Quality Evaluation using Non-Invasive Advanced Imaging Technologies and Computer Vision

Offentlig prosjektsammendrag

The appearance of meat, is important because it is one of the only criteria that a consumer can use to judge the acceptability of the product at the point of purchase. The consumer has expectations of how the lean portion should look i.e., it should be bright in colour and red or pink rather than brown, purple, or grey. How fat is distributed is also important to the consumer, they not only look at the colour of the fat but also how it is distributed and whether the fat is intermuscular or intramuscular. Meat colour is also an important technological quality, which will be favourably correlated to the better water holding capacity of the meat. Consumers use a predetermined view of what they expect a meat product to look like to decide which item to purchase. There are various stages within a supply chain that can affect the final appearance of meat products, however, the most important starting point is the genetics of the animals since this determines the ultimate potential of a product. Genetics companies such as Norsvin drive change by continually assessing the phenotypic values of their genetic stock. For the assessment of the colour and fat content of meat there are two approaches: Subjective assessment and objective assessment. There are two widely recognized methods of subjective assessment for meat, the Japanese colour grading system and the American Colour and Marbling Assessment Procedure (NCCP). Both systems produce data in a discrete format and don’t allow for the interpretation of minor differences. Furthermore, fat composition of the meat is assessed using either Gas Chromatography or Near Infra-Red Spectroscopy (NIR) both of which are destructive methods and don’t accurately interpret the appearance of fat to the consumer nor how it is distributed. Therefore, we aim in this project to develop a meat quality evaluation pipeline using non-invasive multispectral imaging technology and machine learning, based on meat colour and marbling.