This master thesis document aims at improving the accuracy of a painting robot system using vision tracking technologies. The current system does not fulfill industrial expectation in terms of accuracy and occlusion robustness.
A solution is proposed using the Unscented Kalman Filter and a Single-Constraint-At-A-Time approach. The system is first entirely described and methods are analyzed. Then, the SCAAT approach and the Unscented Kalman Filter are described.