FIXATION SIMPLIFIES 3D MOTION ESTIMATION

Fixation is defined as holding the gaze direction towards the same env ironmental point through time. It was proven in the past that fixation reduces the number of unknowns in passive visual navigation from five to four, In this paper, we show that fixation further simplifies 3D m otion estimation because it decouples the motion parameter space, We p roject the spherical motion field in two latitudinal directions with r espect to two different poles of the image sphere. The first projectio n enables the computation of the longitude of the translation directio n and the torsion, The second projection gives the angle between the d irection of fixation and translation, as well as the time to collision to the fixated scene point. Both computational steps are based on one -dimensional searches along meridians of the image sphere. The observe r may move with all six degrees of freedom. We do not use the efferenc e copy of the fixational rotation of the camera, Performance of the al gorithm is tested on real world sequences with fixation accomplished e ither offline or during the recording using an active camera. A compre hensive review of most theories on advantages of fixation clarifies th e differences to our approach. (C) 1997 Academic Press.