Fiber Textile Classification using a Neural Network based on the Neocortical Model

Abstract

As part of sustainability in the production of textile fibers, it is essential to have an accessible and economical method that allows both rural producers and development organizations to identify and classify specific types of fibers, thus guaranteeing a reliable quality measure. This work focuses on the optimization of an object recognition and classification method based on a neocortical model, with the aim of achieving this purpose. Hierarchical Temporal Memory, inspired by the predictive memory theory of the human brain, uses a tree structure of interconnected nodes that apply a specific set of rules to memorize objects in different orientations. According to this model, algorithms based on human vision mechanisms are implemented to preprocess the input images, highlighting the most relevant visual features (similar to how the human brain does). After applying the proposed optimization techniques, compared to the original method, the experimental results show an improvement in performance and accuracy, maintaining the robustness of the original approach.