Egomotion and the stabilized world

The tasks of moving-object detection and motion-based depth recovery as a problem of sensor fusion in the presence of uncertainty are formulated. Two sensor systems are utilized, one providing information about the local image velocity (specifying a point in image-velocity space), and the other providing information about the camera motion (specifying a line segment in image-velocity space). Mahalanobis distance is utilized as a threshold rule for determining consistency between the measurements from the two sensor systems. A framework is suggested for using the resulting segmented flow field to update estimates of the egomotion parameters.