Distance metric learning for feature-agnostic place recognition

Zetao Chen, Stephanie Lowry, Adam Jacobson, Zongyuan Ge, Michael Milford

Research output: Chapter in Book/Report/Conference proceedingConference PaperResearchpeer-review

Abstract

The recent focus on performing visual navigation and place recognition in changing environments has resulted in a large number of heterogeneous techniques each utilizing their own learnt or hand crafted visual features. This paper presents a generally applicable method for learning the appropriate distance metric by which to compare feature responses from any of these techniques in order to perform place recognition under changing environmental conditions. We implement an approach which learns to cluster images captured at spatially proximal locations under different conditions, separated from frames captured at different places. The formulation is a convex optimization, guaranteeing the existence of a global solution. We evaluate the general applicability of our method on two benchmark change datasets using three typical image pre-processing and feature types: GIST, Principal Component Analysis and learnt Convolutional Neural Network features. The results demonstrate that the distance metric learning approach uniformly improves single-image-based visual place recognition performance across all feature types. Furthermore, we demonstrate that this performance improvement is maintained when the sequence-based algorithm SeqSLAM is applied to the single-image place recognition results, leading to state-of-the-art performance.

Original languageEnglish
Title of host publicationIEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2015)
EditorsAlois Knoll
Place of PublicationPiscataway NJ USA
PublisherIEEE, Institute of Electrical and Electronics Engineers
Pages2556-2563
Number of pages8
ISBN (Electronic)9781479999941
ISBN (Print)9781479999958
DOIs
Publication statusPublished - 11 Dec 2015
Externally publishedYes
EventIEEE/RSJ International Conference on Intelligent Robots and Systems 2015 - Hamburg, Germany
Duration: 28 Sep 20152 Oct 2015

Conference

ConferenceIEEE/RSJ International Conference on Intelligent Robots and Systems 2015
Abbreviated titleIROS 2015
CountryGermany
CityHamburg
Period28/09/152/10/15

Keywords

  • Feature extraction
  • Image recognition
  • Lighting
  • Measurement
  • Principal component analysis
  • Training
  • Visualization

Cite this

Chen, Z., Lowry, S., Jacobson, A., Ge, Z., & Milford, M. (2015). Distance metric learning for feature-agnostic place recognition. In A. Knoll (Ed.), IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2015) (pp. 2556-2563). [7353725] Piscataway NJ USA: IEEE, Institute of Electrical and Electronics Engineers. https://doi.org/10.1109/IROS.2015.7353725
Chen, Zetao ; Lowry, Stephanie ; Jacobson, Adam ; Ge, Zongyuan ; Milford, Michael. / Distance metric learning for feature-agnostic place recognition. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2015). editor / Alois Knoll. Piscataway NJ USA : IEEE, Institute of Electrical and Electronics Engineers, 2015. pp. 2556-2563
@inproceedings{cefc9c1fb4ca4dbc93bcc247e244d40f,
title = "Distance metric learning for feature-agnostic place recognition",
abstract = "The recent focus on performing visual navigation and place recognition in changing environments has resulted in a large number of heterogeneous techniques each utilizing their own learnt or hand crafted visual features. This paper presents a generally applicable method for learning the appropriate distance metric by which to compare feature responses from any of these techniques in order to perform place recognition under changing environmental conditions. We implement an approach which learns to cluster images captured at spatially proximal locations under different conditions, separated from frames captured at different places. The formulation is a convex optimization, guaranteeing the existence of a global solution. We evaluate the general applicability of our method on two benchmark change datasets using three typical image pre-processing and feature types: GIST, Principal Component Analysis and learnt Convolutional Neural Network features. The results demonstrate that the distance metric learning approach uniformly improves single-image-based visual place recognition performance across all feature types. Furthermore, we demonstrate that this performance improvement is maintained when the sequence-based algorithm SeqSLAM is applied to the single-image place recognition results, leading to state-of-the-art performance.",
keywords = "Feature extraction, Image recognition, Lighting, Measurement, Principal component analysis, Training, Visualization",
author = "Zetao Chen and Stephanie Lowry and Adam Jacobson and Zongyuan Ge and Michael Milford",
year = "2015",
month = "12",
day = "11",
doi = "10.1109/IROS.2015.7353725",
language = "English",
isbn = "9781479999958",
pages = "2556--2563",
editor = "Alois Knoll",
booktitle = "IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2015)",
publisher = "IEEE, Institute of Electrical and Electronics Engineers",
address = "United States of America",

}

Chen, Z, Lowry, S, Jacobson, A, Ge, Z & Milford, M 2015, Distance metric learning for feature-agnostic place recognition. in A Knoll (ed.), IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2015)., 7353725, IEEE, Institute of Electrical and Electronics Engineers, Piscataway NJ USA, pp. 2556-2563, IEEE/RSJ International Conference on Intelligent Robots and Systems 2015, Hamburg, Germany, 28/09/15. https://doi.org/10.1109/IROS.2015.7353725

Distance metric learning for feature-agnostic place recognition. / Chen, Zetao; Lowry, Stephanie; Jacobson, Adam; Ge, Zongyuan; Milford, Michael.

IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2015). ed. / Alois Knoll. Piscataway NJ USA : IEEE, Institute of Electrical and Electronics Engineers, 2015. p. 2556-2563 7353725.

Research output: Chapter in Book/Report/Conference proceedingConference PaperResearchpeer-review

TY - GEN

T1 - Distance metric learning for feature-agnostic place recognition

AU - Chen, Zetao

AU - Lowry, Stephanie

AU - Jacobson, Adam

AU - Ge, Zongyuan

AU - Milford, Michael

PY - 2015/12/11

Y1 - 2015/12/11

N2 - The recent focus on performing visual navigation and place recognition in changing environments has resulted in a large number of heterogeneous techniques each utilizing their own learnt or hand crafted visual features. This paper presents a generally applicable method for learning the appropriate distance metric by which to compare feature responses from any of these techniques in order to perform place recognition under changing environmental conditions. We implement an approach which learns to cluster images captured at spatially proximal locations under different conditions, separated from frames captured at different places. The formulation is a convex optimization, guaranteeing the existence of a global solution. We evaluate the general applicability of our method on two benchmark change datasets using three typical image pre-processing and feature types: GIST, Principal Component Analysis and learnt Convolutional Neural Network features. The results demonstrate that the distance metric learning approach uniformly improves single-image-based visual place recognition performance across all feature types. Furthermore, we demonstrate that this performance improvement is maintained when the sequence-based algorithm SeqSLAM is applied to the single-image place recognition results, leading to state-of-the-art performance.

AB - The recent focus on performing visual navigation and place recognition in changing environments has resulted in a large number of heterogeneous techniques each utilizing their own learnt or hand crafted visual features. This paper presents a generally applicable method for learning the appropriate distance metric by which to compare feature responses from any of these techniques in order to perform place recognition under changing environmental conditions. We implement an approach which learns to cluster images captured at spatially proximal locations under different conditions, separated from frames captured at different places. The formulation is a convex optimization, guaranteeing the existence of a global solution. We evaluate the general applicability of our method on two benchmark change datasets using three typical image pre-processing and feature types: GIST, Principal Component Analysis and learnt Convolutional Neural Network features. The results demonstrate that the distance metric learning approach uniformly improves single-image-based visual place recognition performance across all feature types. Furthermore, we demonstrate that this performance improvement is maintained when the sequence-based algorithm SeqSLAM is applied to the single-image place recognition results, leading to state-of-the-art performance.

KW - Feature extraction

KW - Image recognition

KW - Lighting

KW - Measurement

KW - Principal component analysis

KW - Training

KW - Visualization

UR - http://www.scopus.com/inward/record.url?scp=84958163306&partnerID=8YFLogxK

U2 - 10.1109/IROS.2015.7353725

DO - 10.1109/IROS.2015.7353725

M3 - Conference Paper

SN - 9781479999958

SP - 2556

EP - 2563

BT - IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2015)

A2 - Knoll, Alois

PB - IEEE, Institute of Electrical and Electronics Engineers

CY - Piscataway NJ USA

ER -

Chen Z, Lowry S, Jacobson A, Ge Z, Milford M. Distance metric learning for feature-agnostic place recognition. In Knoll A, editor, IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2015). Piscataway NJ USA: IEEE, Institute of Electrical and Electronics Engineers. 2015. p. 2556-2563. 7353725 https://doi.org/10.1109/IROS.2015.7353725