método e sistema para reconhecimento e classificação de objetos em ambientes semi controlados

The present invention relates to a method, namely a system for recognition and categorization of objects (e.g. fruit, legumes, vegetables, flowers, boxes, etc.) in semi-controlled environments, where there is no use of diverse sources of light of different wavelengths (e.g. supermarkets, fairs, distribution centres). The present technology allows the extraction of descriptive characteristics of a dataset (e.g. images of the objects to be categorized) and combination thereof for the purpose of generating a more robust and flexible classifier. This invention relates to the combination of different types of descriptors and also the association of various types of classifiers, each specialized and better configured for a set of characteristics or types in a confused classification. There is an intelligent combination of binary classifiers that divide the greater multi-type classification problem into small problems. The present invention is capable of aggregating the information extracted from various types of media with a view to solving the problem of classifying objects in internal environments, such as a distribution centre or a supermarket, with the aim of automating the purchasing process.