Primerjava metod za kalibracijo kamere in robotskega manipulatorja

GRČAR, FILIP (author), Skočaj, Danijel (mentor)

Kalibracija kamere na robotskem manipulatorju je splošen problem ugotavljanja lege togo nameščene kamere na robotskem manipulatorju glede na vrh robota. Običajno se izvaja z zajemanjem slik fiksnega predmeta znanih dimenzij, pri čemer je robotski manipulator za vsako sliko drugače pozicioniran. V tej diplomski nalogi primerjamo kalibracijo na dveh sodelujočih robotih UR10e in Aubo i5 v kombinaciji s tremi kamerami Zivid One+ Medium, Intel Realsense D435 in predstavnikom kamer Basler Ace. Na danih kombinacijah smo preizkusili pet oz. šest različnih metod kalibracije v odvisnosti od števila zajetih slik. Ugotovili smo, da se natančnost kalibracije s povečevanjem števila slik povečuje, najbolje pa sta se odrezali metodi Parka in Horauda. Med kamerami se je najbolje izkazal Baslerjev produkt, med roboti pa po tehtnem premisleku UR10e.

• Language: Slovenian
• Keywords: kalibracija, sodelujoči roboti, ROS, OpenCV
• Work type: Bachelor thesis/paper
• Typology: 2.11 - Undergraduate Thesis
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2020
• PID: 20.500.12556/RUL-121029
• COBISS.SI-ID: 33070339
• Publication date in RUL: 29.09.2020

Secondary language

• Language: English
• Title: Comparison of Robot Hand-Eye Calibration Methods

Abstract

Hand-eye calibration is a well known problem of defining a robot-mounted camera's pose relative to the robot's end-effector. This is usually achieved by capturing multiple images of a static object of known dimensions where the robot manipulator is differently positioned for each image. In this thesis we compare two collaborative robots UR10e and Aubo i5 in combination with three cameras Zivid One+ Medium, Intel Realsense D435 and one of the Basler's Ace camera family member. We compared five to six different calibration methods on all of mentioned robot-camera combinations and its dependencies on images' number. We found out that the accuracy of the methods rises with the number of the images and the best methods turned out to be Park's and Horaud's. Among the cameras the best turned out to be the Basler's one and the best robot after some extra cosideration became UR10e.

• Keywords: hand-eye calibration, collaborative robots, ROS, OpenCV