Fiber recognition and classification optimizing a neural network based on the neocortical model

Abstract

In the productive sustainability of animal textil fiber framework, having a quick, safe and inexpensive method that allows to obtain a measure of fiber quality to compete in local and international markets, is highly beneficial for farmers. In this work, a method of recognition and classification of objects based on a Hierarchical Temporary Memory is optimized for that purpose. The Hierarchical Temporary Memory, inspired by the theory of prediction memory of the human brain, consists of a tree structure of computationally connected nodes which have a p articular set of rules for memorizing objects that appear in different orientations [Hawkins & Ahmad]. The input images were subjected to a preprocessing based on a mathematical model to highlight the most relevant visual characteristics. The experimental results demonstrated an improvement in performance and precision.