Merging multi-camera data to reduce motion analysis instrumental errors using Kalman filters

In motion capture systems, markers are often seen by multiple cameras. All cameras do not measure the position of the markers with the same reliability because of environmental factors such as the position of the marker in the field of view or the light intensity received by the cameras. Kalman filters offer a general framework to take the reliability of the various cameras into account and consequently improve the estimation of the marker position. The proposed process can be applied to both passive and active systems. Several reliability models of the cameras are compared for the Codamotion active system, which is considered as a specific illustration. The proposed method significantly reduces the noise in the signal, especially at long-range distances. Therefore, it improves the confidence of the positions at the limits of the field of view.