Robust structure from motion under weak perspective

Levente Hajder, D. Chetverikov, István Vajk

Research output: Chapter in Book/Report/Conference proceedingConference contribution

4 Citations (Scopus)

Abstract

It is widely known that, for the affine camera model, both shape and motion data can be factorized directly from the measurement matrix constructed from 2D image points coordinates. However, classical algorithms for Structure from Motion (SfM) are not robust: measurement outliers, that is, incorrectly detected or matched feature points can destroy the result. A few methods to robustify SfM have already been proposed. Different outlier detection schemes have been used. We examine an efficient algorithm by Trajković et al. who use affine camera model and the Least Median of Squares (LMedS) method to separate inliers from outliers. LMedS is only applicable when the ratio of inliers exceeds 50%. We show that the Least Trimmed Squares (LTS) method is more efficient in robust SfM than LMedS. In particular, we demonstrate that LTS can handle inlier ratios below 50%. We also show that using the real (Euclidean) motion data results in a more precise SfM algorithm that using the affine camera model. Based on these observations, we propose a novel robust SfM algorithm and discuss its advantages and limits. The proposed method and the Trajkovic procedure are quantitatively compared on synthetic data in different simulated situations. The methods are also tested on synthesized and real video sequences.

Original languageEnglish
Title of host publicationProceedings - 2nd International Symposium on 3D Data Processing, Visualization, and Transmission. 3DPVT 2004
EditorsY. Aloimonos, G. Taubin
Pages828-835
Number of pages8
DOIs
Publication statusPublished - 2004
EventProceedings - 2nd International Symposium on 3D Data Processing, Visualization, and Transmission. 3DPVT 2004 - Thessaloniki, Greece
Duration: Sep 6 2004Sep 9 2004

Other

OtherProceedings - 2nd International Symposium on 3D Data Processing, Visualization, and Transmission. 3DPVT 2004
CountryGreece
CityThessaloniki
Period9/6/049/9/04

Fingerprint

Cameras

ASJC Scopus subject areas

  • Engineering(all)

Cite this

Hajder, L., Chetverikov, D., & Vajk, I. (2004). Robust structure from motion under weak perspective. In Y. Aloimonos, & G. Taubin (Eds.), Proceedings - 2nd International Symposium on 3D Data Processing, Visualization, and Transmission. 3DPVT 2004 (pp. 828-835) https://doi.org/10.1109/TDPVT.2004.1335401

Robust structure from motion under weak perspective. / Hajder, Levente; Chetverikov, D.; Vajk, István.

Proceedings - 2nd International Symposium on 3D Data Processing, Visualization, and Transmission. 3DPVT 2004. ed. / Y. Aloimonos; G. Taubin. 2004. p. 828-835.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Hajder, L, Chetverikov, D & Vajk, I 2004, Robust structure from motion under weak perspective. in Y Aloimonos & G Taubin (eds), Proceedings - 2nd International Symposium on 3D Data Processing, Visualization, and Transmission. 3DPVT 2004. pp. 828-835, Proceedings - 2nd International Symposium on 3D Data Processing, Visualization, and Transmission. 3DPVT 2004, Thessaloniki, Greece, 9/6/04. https://doi.org/10.1109/TDPVT.2004.1335401
Hajder L, Chetverikov D, Vajk I. Robust structure from motion under weak perspective. In Aloimonos Y, Taubin G, editors, Proceedings - 2nd International Symposium on 3D Data Processing, Visualization, and Transmission. 3DPVT 2004. 2004. p. 828-835 https://doi.org/10.1109/TDPVT.2004.1335401
Hajder, Levente ; Chetverikov, D. ; Vajk, István. / Robust structure from motion under weak perspective. Proceedings - 2nd International Symposium on 3D Data Processing, Visualization, and Transmission. 3DPVT 2004. editor / Y. Aloimonos ; G. Taubin. 2004. pp. 828-835
@inproceedings{66f96606fde74f67ae54cac852fe696e,
title = "Robust structure from motion under weak perspective",
abstract = "It is widely known that, for the affine camera model, both shape and motion data can be factorized directly from the measurement matrix constructed from 2D image points coordinates. However, classical algorithms for Structure from Motion (SfM) are not robust: measurement outliers, that is, incorrectly detected or matched feature points can destroy the result. A few methods to robustify SfM have already been proposed. Different outlier detection schemes have been used. We examine an efficient algorithm by Trajković et al. who use affine camera model and the Least Median of Squares (LMedS) method to separate inliers from outliers. LMedS is only applicable when the ratio of inliers exceeds 50{\%}. We show that the Least Trimmed Squares (LTS) method is more efficient in robust SfM than LMedS. In particular, we demonstrate that LTS can handle inlier ratios below 50{\%}. We also show that using the real (Euclidean) motion data results in a more precise SfM algorithm that using the affine camera model. Based on these observations, we propose a novel robust SfM algorithm and discuss its advantages and limits. The proposed method and the Trajkovic procedure are quantitatively compared on synthetic data in different simulated situations. The methods are also tested on synthesized and real video sequences.",
author = "Levente Hajder and D. Chetverikov and Istv{\'a}n Vajk",
year = "2004",
doi = "10.1109/TDPVT.2004.1335401",
language = "English",
isbn = "0769522238",
pages = "828--835",
editor = "Y. Aloimonos and G. Taubin",
booktitle = "Proceedings - 2nd International Symposium on 3D Data Processing, Visualization, and Transmission. 3DPVT 2004",

}

TY - GEN

T1 - Robust structure from motion under weak perspective

AU - Hajder, Levente

AU - Chetverikov, D.

AU - Vajk, István

PY - 2004

Y1 - 2004

N2 - It is widely known that, for the affine camera model, both shape and motion data can be factorized directly from the measurement matrix constructed from 2D image points coordinates. However, classical algorithms for Structure from Motion (SfM) are not robust: measurement outliers, that is, incorrectly detected or matched feature points can destroy the result. A few methods to robustify SfM have already been proposed. Different outlier detection schemes have been used. We examine an efficient algorithm by Trajković et al. who use affine camera model and the Least Median of Squares (LMedS) method to separate inliers from outliers. LMedS is only applicable when the ratio of inliers exceeds 50%. We show that the Least Trimmed Squares (LTS) method is more efficient in robust SfM than LMedS. In particular, we demonstrate that LTS can handle inlier ratios below 50%. We also show that using the real (Euclidean) motion data results in a more precise SfM algorithm that using the affine camera model. Based on these observations, we propose a novel robust SfM algorithm and discuss its advantages and limits. The proposed method and the Trajkovic procedure are quantitatively compared on synthetic data in different simulated situations. The methods are also tested on synthesized and real video sequences.

AB - It is widely known that, for the affine camera model, both shape and motion data can be factorized directly from the measurement matrix constructed from 2D image points coordinates. However, classical algorithms for Structure from Motion (SfM) are not robust: measurement outliers, that is, incorrectly detected or matched feature points can destroy the result. A few methods to robustify SfM have already been proposed. Different outlier detection schemes have been used. We examine an efficient algorithm by Trajković et al. who use affine camera model and the Least Median of Squares (LMedS) method to separate inliers from outliers. LMedS is only applicable when the ratio of inliers exceeds 50%. We show that the Least Trimmed Squares (LTS) method is more efficient in robust SfM than LMedS. In particular, we demonstrate that LTS can handle inlier ratios below 50%. We also show that using the real (Euclidean) motion data results in a more precise SfM algorithm that using the affine camera model. Based on these observations, we propose a novel robust SfM algorithm and discuss its advantages and limits. The proposed method and the Trajkovic procedure are quantitatively compared on synthetic data in different simulated situations. The methods are also tested on synthesized and real video sequences.

UR - http://www.scopus.com/inward/record.url?scp=16244394059&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=16244394059&partnerID=8YFLogxK

U2 - 10.1109/TDPVT.2004.1335401

DO - 10.1109/TDPVT.2004.1335401

M3 - Conference contribution

AN - SCOPUS:16244394059

SN - 0769522238

SN - 9780769522234

SP - 828

EP - 835

BT - Proceedings - 2nd International Symposium on 3D Data Processing, Visualization, and Transmission. 3DPVT 2004

A2 - Aloimonos, Y.

A2 - Taubin, G.

ER -