Expand this Topic clickable element to expand a topic
Skip to content
Optica Publishing Group

Egomotion without depth estimation

Open Access Open Access

Abstract

The motion field for an observer moving through a static environment depends on the observer's translational and rotational velocities along with the distances to surface points. We show how the equations describing the motion field can be split by an exact algebraic manipulation to form three sets of equations, the first of which relates properties of the motion field to the observer's direction of translational motion alone. Therefore, to solve for the direction of translation, the depths and rotational velocity need not be known or estimated. Unlike previous surface-based techniques, we require no assumptions about the smoothness of the surface from which the optic flow measurements are taken, or that the measurements come from the same surface. Instead, we show that this procedure can provide a robust estimate for the direction of translation, given simply a sparse sampling of the optic flow from scenes that contain a significant variation in depth. Moreover, given this estimate we show how two linear equations can be found, one of which provides an overdetermined system of the rotational velocity alone, and the other contains only the relative depths. Again we show that the rotational velocity can be computed robustly, given a sparse sampling of velocities.

© 1989 Optical Society of America

PDF Article
More Like This
Egomotion without correspondence

Berthold K. P. Horn
MA1 OSA Annual Meeting (FIO) 1989

Shape from Egomotion: A Model-Based Probabilistic Approach

Yi-Ping Hung and David B. Cooper
MC2 Image Understanding and Machine Vision (IUMV) 1989

Motion Estimation Using Surface Correspondences*

Bikash Sabata, N. Nandhakumar, and J.K. Aggarwal
TuA1 Image Understanding and Machine Vision (IUMV) 1989

Select as filters


Select Topics Cancel
© Copyright 2024 | Optica Publishing Group. All rights reserved, including rights for text and data mining and training of artificial technologies or similar technologies.