Sistema para reconhecimento de objetos voltados para aplicações de realidade aumentada na indústria

Abstract

This course conclusion work presents an object recognition software project, aimed at augmented reality applications in the industry. The system will be developed for the Android platform within Ubivis, a Curitiba based company that develops software industrial processes optimization. This project is part of a larger work developed at the company that aims to build a knowledge base for augmented reality applications, such as support in machine data visualization for operations, visualization of machine instructions for maintenance personnel and remote assistance for field auditors. The objective of the work proposed here is to develop an experimentally viable base for object recognition application that may be extended to specific domains. The initially proposed system will consist mainly of a module for object identification, whose architecture will be based on neural networks with the TensorFlow framework. For the proposal validation as a whole, a mobile application will be developed for industry.