Özer, N.Erdem, TanjuErcan, Ali ÖzerEroğlu Erdem, Ç.2016-02-162016-02-162012978-1-4673-0054-42165-0608http://hdl.handle.net/10679/2548https://doi.org/10.1109/SIU.2012.6204725Due to copyright restrictions, the access to the full text of this article is only available via subscription.It is well known in a Bayesian filtering framework, the use of inertial sensors such as accelerometers and gyroscopes improves 3D tracking performance compared to using camera measurements only. The performance improvement is more evident when the camera undergoes a high degree of motion. However, it is not well known whether the inertial sensors should be used as control inputs or as measurements. In this paper, we present the results of an extensive set of simulations comparing different combinations of using inertial sensors as control inputs or as measurements. We show that it is better use a gyroscope as a control input while an accelerometer can be used as a measurement or control input. We also derive and present the extended Kalman filter (EKF) equations for a specific case of fusing accelerometer and gyroscope data that has not been reported before.turinfo:eu-repo/semantics/restrictedAccessConference paper1410.1109/SIU.2012.6204725Bayes methodsKalman filtersAccelerometersCamerasGyroscopesInertial systemsNonlinear filtersSensor fusion2-s2.0-84863449596