Sistema de visão robótica aplicado a tarefas de manipulação

Abstract

This work presents a robotic vision system applied to part manipulation tasks. The system developed is capable of recognizing, locating and manipulating a single part located in a position initially unknown to it. A system was created, composed of a robot characterized by a robotic arm that manipulates the part, a camera to capture images of the part in its visual field, which is delimited by the surface of a table built for this purpose, and a computer to process the images captured by the camera and send instructions to the robot. Therefore, the project makes use of the Vision Builder and LabVIEW softwares, both developed by National Instruments. The system offers flexibility in relation to the position of the robot and the table - there’s no need of both being aligned or parallel, as the part can be placed anywhere inside the reach of the robot and the visual field of the camera. Lab tests were made in order to test the proper operation of the robotic vision system. The tests were performed from an initial calibration, where the part is placed in two pre-established positions and after that point, the robot visualizes and manipulates the part. In order to execute these operations, topography and geodesy concepts were used, due to the vision system being in a different coordinate system than the one originated by the camera and the variation on the position of both the robot and the table. To assure the compatibility of the coordinate systems, the cited concepts were required, as to make rotations and translations, as well as to compensate errors in the measurements acquired. Therefore, this project can have important applications in the automation industry, considering its flexibility, since the robot will not only act based on manually input instructions, but also on perceptions of change in its field of reach through its robotic vision.