Visual odometry: Part II: Matching, robustness, optimization, and applications

Friedrich Fraundorfer, Davide Scaramuzza

Research output: Contribution to journalArticleResearchpeer-review

Abstract

Visual odometry (VO) is the process of estimating the gomotion of an agent using the input of a single or multiple cameras attached to it. The advantage of VO with respect to wheel odometry is that VO is not affected by wheel slip in uneven terrain or other adverse conditions. During the feature-detection step, the image is searched for salient keypoints that are likely to match well in other images. A local feature is an image pattern that differs from its immediate neighborhood in terms of intensity, color, and texture. In the feature description step, the region around each detected feature is converted into a compact descriptor that can be matched against other descriptors. After comparing all feature descriptors between two images, the best correspondence of a feature in the second image is chosen as that with the closest descriptor. Alternatively, if only the motion model is known but not the 3-D feature position, the corresponding match can be searched along the epipolar line in the second image.

Original languageEnglish
Article number6153423
Pages (from-to)78-90
Number of pages13
JournalIEEE Robotics & Automation Magazine
Volume19
Issue number2
DOIs
Publication statusPublished - 2012

Fingerprint

Wheels
Textures
Cameras
Color

ASJC Scopus subject areas

  • Electrical and Electronic Engineering
  • Control and Systems Engineering
  • Computer Science Applications

Cite this

Visual odometry : Part II: Matching, robustness, optimization, and applications. / Fraundorfer, Friedrich; Scaramuzza, Davide.

In: IEEE Robotics & Automation Magazine , Vol. 19, No. 2, 6153423, 2012, p. 78-90.

Research output: Contribution to journalArticleResearchpeer-review

@article{65888dc89cbf46b180b6a6bbd91f5828,
title = "Visual odometry: Part II: Matching, robustness, optimization, and applications",
abstract = "Visual odometry (VO) is the process of estimating the gomotion of an agent using the input of a single or multiple cameras attached to it. The advantage of VO with respect to wheel odometry is that VO is not affected by wheel slip in uneven terrain or other adverse conditions. During the feature-detection step, the image is searched for salient keypoints that are likely to match well in other images. A local feature is an image pattern that differs from its immediate neighborhood in terms of intensity, color, and texture. In the feature description step, the region around each detected feature is converted into a compact descriptor that can be matched against other descriptors. After comparing all feature descriptors between two images, the best correspondence of a feature in the second image is chosen as that with the closest descriptor. Alternatively, if only the motion model is known but not the 3-D feature position, the corresponding match can be searched along the epipolar line in the second image.",
author = "Friedrich Fraundorfer and Davide Scaramuzza",
year = "2012",
doi = "10.1109/MRA.2012.2182810",
language = "English",
volume = "19",
pages = "78--90",
journal = "IEEE Robotics & Automation Magazine",
issn = "1070-9932",
publisher = "IEEE Robotics and Automation Society",
number = "2",

}

TY - JOUR

T1 - Visual odometry

T2 - Part II: Matching, robustness, optimization, and applications

AU - Fraundorfer, Friedrich

AU - Scaramuzza, Davide

PY - 2012

Y1 - 2012

N2 - Visual odometry (VO) is the process of estimating the gomotion of an agent using the input of a single or multiple cameras attached to it. The advantage of VO with respect to wheel odometry is that VO is not affected by wheel slip in uneven terrain or other adverse conditions. During the feature-detection step, the image is searched for salient keypoints that are likely to match well in other images. A local feature is an image pattern that differs from its immediate neighborhood in terms of intensity, color, and texture. In the feature description step, the region around each detected feature is converted into a compact descriptor that can be matched against other descriptors. After comparing all feature descriptors between two images, the best correspondence of a feature in the second image is chosen as that with the closest descriptor. Alternatively, if only the motion model is known but not the 3-D feature position, the corresponding match can be searched along the epipolar line in the second image.

AB - Visual odometry (VO) is the process of estimating the gomotion of an agent using the input of a single or multiple cameras attached to it. The advantage of VO with respect to wheel odometry is that VO is not affected by wheel slip in uneven terrain or other adverse conditions. During the feature-detection step, the image is searched for salient keypoints that are likely to match well in other images. A local feature is an image pattern that differs from its immediate neighborhood in terms of intensity, color, and texture. In the feature description step, the region around each detected feature is converted into a compact descriptor that can be matched against other descriptors. After comparing all feature descriptors between two images, the best correspondence of a feature in the second image is chosen as that with the closest descriptor. Alternatively, if only the motion model is known but not the 3-D feature position, the corresponding match can be searched along the epipolar line in the second image.

UR - http://www.scopus.com/inward/record.url?scp=84862585064&partnerID=8YFLogxK

U2 - 10.1109/MRA.2012.2182810

DO - 10.1109/MRA.2012.2182810

M3 - Article

VL - 19

SP - 78

EP - 90

JO - IEEE Robotics & Automation Magazine

JF - IEEE Robotics & Automation Magazine

SN - 1070-9932

IS - 2

M1 - 6153423

ER -