Neural mapping and parallel optical flow computation for autonomous navigation

Heinrich Bulthoff, James J. Little, Hanspeter A. Mallot

Research output: Contribution to journalArticle

Abstract

We present two information processing strategies, derived from neurobiology, which facilitate the evaluation of optical flow data considerably. One is a parallel motion algorithm, and the other is an inverse perspective mapping technique. The combination of the two implements the following principles of biological information processing: (1) Spatially dense motion data are obtained which are not biased by the aperture problem. (2) The computational resources available can be utilized most effectively by transforming space-variant problems into space-invariant ones. (3) The definition of an obstacle is reduced to its most basic meaning: it is only the elevation above the ground plane that leads to a detection and pattern recognition is not necessary at this stage. This integrated approach has been successfully tested on real-time robot navigation applications.

Original languageEnglish
JournalNeural Networks
Volume1
Issue number1 SUPPL
DOIs
Publication statusPublished - 1988 Dec 1
Externally publishedYes

Fingerprint

Optical flows
Automatic Data Processing
Navigation
Neurobiology
Pattern recognition
Robots

ASJC Scopus subject areas

  • Artificial Intelligence
  • Neuroscience(all)

Cite this

Neural mapping and parallel optical flow computation for autonomous navigation. / Bulthoff, Heinrich; Little, James J.; Mallot, Hanspeter A.

In: Neural Networks, Vol. 1, No. 1 SUPPL, 01.12.1988.

Research output: Contribution to journalArticle

Bulthoff, Heinrich ; Little, James J. ; Mallot, Hanspeter A. / Neural mapping and parallel optical flow computation for autonomous navigation. In: Neural Networks. 1988 ; Vol. 1, No. 1 SUPPL.
@article{2598242dbe834b9994001ebb01602556,
title = "Neural mapping and parallel optical flow computation for autonomous navigation",
abstract = "We present two information processing strategies, derived from neurobiology, which facilitate the evaluation of optical flow data considerably. One is a parallel motion algorithm, and the other is an inverse perspective mapping technique. The combination of the two implements the following principles of biological information processing: (1) Spatially dense motion data are obtained which are not biased by the aperture problem. (2) The computational resources available can be utilized most effectively by transforming space-variant problems into space-invariant ones. (3) The definition of an obstacle is reduced to its most basic meaning: it is only the elevation above the ground plane that leads to a detection and pattern recognition is not necessary at this stage. This integrated approach has been successfully tested on real-time robot navigation applications.",
author = "Heinrich Bulthoff and Little, {James J.} and Mallot, {Hanspeter A.}",
year = "1988",
month = "12",
day = "1",
doi = "10.1016/0893-6080(88)90504-7",
language = "English",
volume = "1",
journal = "Neural Networks",
issn = "0893-6080",
publisher = "Elsevier Limited",
number = "1 SUPPL",

}

TY - JOUR

T1 - Neural mapping and parallel optical flow computation for autonomous navigation

AU - Bulthoff, Heinrich

AU - Little, James J.

AU - Mallot, Hanspeter A.

PY - 1988/12/1

Y1 - 1988/12/1

N2 - We present two information processing strategies, derived from neurobiology, which facilitate the evaluation of optical flow data considerably. One is a parallel motion algorithm, and the other is an inverse perspective mapping technique. The combination of the two implements the following principles of biological information processing: (1) Spatially dense motion data are obtained which are not biased by the aperture problem. (2) The computational resources available can be utilized most effectively by transforming space-variant problems into space-invariant ones. (3) The definition of an obstacle is reduced to its most basic meaning: it is only the elevation above the ground plane that leads to a detection and pattern recognition is not necessary at this stage. This integrated approach has been successfully tested on real-time robot navigation applications.

AB - We present two information processing strategies, derived from neurobiology, which facilitate the evaluation of optical flow data considerably. One is a parallel motion algorithm, and the other is an inverse perspective mapping technique. The combination of the two implements the following principles of biological information processing: (1) Spatially dense motion data are obtained which are not biased by the aperture problem. (2) The computational resources available can be utilized most effectively by transforming space-variant problems into space-invariant ones. (3) The definition of an obstacle is reduced to its most basic meaning: it is only the elevation above the ground plane that leads to a detection and pattern recognition is not necessary at this stage. This integrated approach has been successfully tested on real-time robot navigation applications.

UR - http://www.scopus.com/inward/record.url?scp=0024174089&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0024174089&partnerID=8YFLogxK

U2 - 10.1016/0893-6080(88)90504-7

DO - 10.1016/0893-6080(88)90504-7

M3 - Article

AN - SCOPUS:0024174089

VL - 1

JO - Neural Networks

JF - Neural Networks

SN - 0893-6080

IS - 1 SUPPL

ER -