Traditional marbling meat evaluation is a tedious, repetitive, costly and time-consuming task performed by panellists. Alternatively, we have Computer Vision Systems (CVS) to mitigate these problems. However, most of CVS are restricted to specific environments, configurations or muscle types, and marbling scores are settled for a particular marbling meat standard. In this context, we developed a CVS for meat marbling grading, which is flexible to different muscle colour contrasts and grading standards. Essentially, the proposed method segments an image pre-processed by illumination normalisation and contrast enhancement, analyses visible intramuscular fat pixels and attributes a score based on a desired meat standard defined in the learning step. Learning approach is an instance-based system making use of k-Nearest Neighbours algorithm (k-NN) to attribute a score from segmentation results. The algorithm classifies the new samples based on scores assigned by panellists. We investigated the optimal number of samples for modelling, focusing on the smallest number leading to acceptable accuracy, and considering two different animal species: bovine and swine. The CVS led to accuracy values equal to 81.59% (bovine) and to 76.14% (swine), using only three samples for each marbling score.
