STRATEGY BASED ON ARTIFICIAL VISION AND SOM NET FOR THE CLASSIFICATION OF MARBLED MEAT

Abstract

The classification of marbled meat from the cuts is very important since it automatically allows the identification of the type of meat to which it belongs, generating a tool base for the goat industry for its easy and prompt selection. This classification allows it to be done based on the images in .jpg format of the marbling cuts (being 104 images as input for the selection), considering the different conditions taken for their analysis and classification (Low, medium, high), to achieve said classification, a tool based on artificial intelligence is used, specifically the SOM neural network. In such a way to make the classification easier (according to the quantity, shape and accumulation of intermuscular fat present), since it does not require a specialist or with extensive knowledge in the identification of types of meat, to carry out its classification, placing it in equipment or automated machinery for use in industries. The validation is carried out using the confusion matrix, achieving a sensitivity of 1.0 and a specificity of 0.94 and a precision of 0.83. The strategy for preparing data based on artificial vision, until obtaining data for input to the SOM neural network is detailed step by step in this article