Recognition and simulation of a vision system for a KUKA KR-5 industrial manipulator

摘要

The objective of this article is to have a computer vision system capable of recognizing various objects in a given plan, so that a KUKA KR-5 industrial robot can perform manipulation tasks with said objects. Three algorithms are evaluated to improve the performance of this vision system, with the ability to recognize multiple objects and their positions. A Microsoft Kinect 2.0 sensor is used as an input device to a system with a KUKA KR-5 industrial robotic manipulator. Object recognition is done by two algorithms; FastOrb and Brisk, while a third Ransac-PCA algorithm is used for object detection. With the combination of these algorithms, it is possible to recognize more than two objects in a three-dimensional working environment. Finally, the Moveit algorithm is used to obtain trajectories and calculate the inverse kinematics and Gazebo to obtain the simulation of the KUKA KR-5 industrial manipulator.