Real-Time tracking system using Single-Constraints-At- A-Time and Unscented Kalman Filter
Author
Le Reun, Gauthier
Term
4. term
Publication year
2010
Submitted on
2010-06-03
Pages
64
Abstract
Denne kandidatafhandling undersøger, hvordan en industriel malerobot kan blive mere præcis ved hjælp af kamerabaseret visionsporing. Det nuværende system lever ikke op til industrielle forventninger til nøjagtighed og robusthed over for okklusioner—situationer, hvor kameraets udsyn delvist er blokeret. Afhandlingen foreslår en sporingsløsning, der kombinerer Unscented Kalman Filter (UKF), en statistisk metode til at estimere robotens tilstand ud fra støjfyldte målinger, med en Single-Constraint-At-A-Time (SCAAT)-tilgang, som opdaterer estimater ved én måling ad gangen. Først beskrives systemet og de anvendte metoder analyseres; derefter gennemgås UKF og SCAAT i detaljer.
This thesis explores how to make an industrial painting robot more precise using camera-based vision tracking. The current system does not meet industrial expectations for accuracy or robustness to occlusions—situations where the camera’s view is partly blocked. The work proposes a tracking solution that combines the Unscented Kalman Filter (UKF), a statistical method for estimating the robot’s state from noisy measurements, with a Single-Constraint-At-A-Time (SCAAT) approach, which updates estimates using one measurement at a time. The thesis first describes the system and analyzes the methods, then details UKF and SCAAT.
[This abstract was generated with the help of AI]
Documents