Classification of Solid Objects with Defined Shapes Using Stereoscopic Vision and a Robotic Arm

Abstract

Summary form only given. This project implements a didactic module, which uses stereoscopic vision for depth estimation. The system identifies the XYZ coordinates where the solid objects with defined shapes are located, and then a robotic arm is used to manipulate and classify the objects according to their shape and color. Computational algorithms were developed on LabVIEW <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">®</sup> for image acquisition and control of the robotic arm. MATLAB <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">®</sup> was used to solve the stereoscopic vision problem. Furthermore, an application based on an artificial neural network was trained with color and texture features to identify lemons, apples, oranges, tangerines and tomatoes. In order to verify the system operation, tests were performed, the results show 2% of error in the scene reconstruction and a 10% error in positioning of the robotic arm over the identified fruit.