PINFI - Tool for image classification with artificial neural networks and fuzzy logic

This article presents a modeling tool for image classification using artificial neural networks (RNA) and fuzzy logic, developed with focus on the industrial scenario. The design of this tool started with a previous research of RNA and neuro-fuzzy modeling tools and a survey-type research in which it was identified that the industries of a certain region do not know these technologies or do not know how to apply it in their processes. From this, the tool called Neuro-Fuzzy Image Processor for Industry (PINFI) was developed based on the demands of the survey and integrating open source software libraries researched in the bibliographic survey. This tool allows the modeling of projects in five stages: Acquisition of images; pre-processing; fuzzification block; RNA modeling; and the presentation of the results (output). The PINFI was validated in terms of ergonomics and software usability, meeting 91% of the requirements evaluated. As for the accuracy of the classifier, it reached recognition rates of 93% to 100% in the bases evaluated. These bases included images of tobacco leaves, vehicle wheels and human faces, which demonstrates that PINFI has the potential to be used in industrial demands and also in diverse applications.