Context-Based Filtering of Noisy Labels for Automatic Basemap Updating From UAV Data

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

Abstract

Unmanned aerial vehicles (UAVs) have the potential to obtain high-resolution aerial imagery at frequent intervals, making them a valuable tool for urban planners who require up-to-date basemaps. Supervised classification methods can be exploited to translate the UAV data into such basemaps. However, these methods require labeled training samples, the collection of which may be complex and time consuming. Existing spatial datasets can be exploited to provide the training labels, but these often contain errors due to differences in the date or resolution of the dataset from which these outdated labels were obtained. In this paper, we propose an approach for updating basemaps using global and local contextual cues to automatically remove unreliable samples from the training set, and thereby, improve the classification accuracy. Using UAV datasets over Kigali, Rwanda, and Dar es Salaam, Tanzania, we demonstrate how the amount of mislabeled training samples can be reduced by 44.1% and 35.5%, respectively, leading to a classification accuracy of 92.1% in Kigali and 91.3% in Dar es Salaam. To achieve the same accuracy in Dar es Salaam, between 50000 and 60000 manually labeled image segments would be needed. This demonstrates that the proposed approach of using outdated spatial data to provide labels and iteratively removing unreliable samples is a viable method for obtaining high classification accuracies while reducing the costly step of acquiring labeled training samples.
Original languageEnglish
Title of host publicationIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Pages1-11
Number of pages11
Volume11
Edition8
DOIs
Publication statusPublished - Aug 2018

Publication series

NameIEEE Journal of selected topics in applied earth observations and remote sensing
PublisherIEEE
ISSN (Print)1939-1404

Fingerprint

image classification
spatial data
imagery
method
vehicle

Keywords

  • ITC-ISI-JOURNAL-ARTICLE

Cite this

Gevaert, C. M., Persello, C., Oude Elberink, S. J., Vosselman, G., & Sliuzas, R. V. (2018). Context-Based Filtering of Noisy Labels for Automatic Basemap Updating From UAV Data. In IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing (8 ed., Vol. 11, pp. 1-11). (IEEE Journal of selected topics in applied earth observations and remote sensing). https://doi.org/10.1109/JSTARS.2017.2762905
Gevaert, C.M. ; Persello, C. ; Oude Elberink, S.J. ; Vosselman, G. ; Sliuzas, R.V. / Context-Based Filtering of Noisy Labels for Automatic Basemap Updating From UAV Data. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing. Vol. 11 8. ed. 2018. pp. 1-11 (IEEE Journal of selected topics in applied earth observations and remote sensing).
@inproceedings{f4e8e6cb746c48f1b34b5102a88c96a8,
title = "Context-Based Filtering of Noisy Labels for Automatic Basemap Updating From UAV Data",
abstract = "Unmanned aerial vehicles (UAVs) have the potential to obtain high-resolution aerial imagery at frequent intervals, making them a valuable tool for urban planners who require up-to-date basemaps. Supervised classification methods can be exploited to translate the UAV data into such basemaps. However, these methods require labeled training samples, the collection of which may be complex and time consuming. Existing spatial datasets can be exploited to provide the training labels, but these often contain errors due to differences in the date or resolution of the dataset from which these outdated labels were obtained. In this paper, we propose an approach for updating basemaps using global and local contextual cues to automatically remove unreliable samples from the training set, and thereby, improve the classification accuracy. Using UAV datasets over Kigali, Rwanda, and Dar es Salaam, Tanzania, we demonstrate how the amount of mislabeled training samples can be reduced by 44.1{\%} and 35.5{\%}, respectively, leading to a classification accuracy of 92.1{\%} in Kigali and 91.3{\%} in Dar es Salaam. To achieve the same accuracy in Dar es Salaam, between 50000 and 60000 manually labeled image segments would be needed. This demonstrates that the proposed approach of using outdated spatial data to provide labels and iteratively removing unreliable samples is a viable method for obtaining high classification accuracies while reducing the costly step of acquiring labeled training samples.",
keywords = "ITC-ISI-JOURNAL-ARTICLE",
author = "C.M. Gevaert and C. Persello and {Oude Elberink}, S.J. and G. Vosselman and R.V. Sliuzas",
year = "2018",
month = "8",
doi = "10.1109/JSTARS.2017.2762905",
language = "English",
volume = "11",
series = "IEEE Journal of selected topics in applied earth observations and remote sensing",
publisher = "IEEE",
pages = "1--11",
booktitle = "IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing",
edition = "8",

}

Gevaert, CM, Persello, C, Oude Elberink, SJ, Vosselman, G & Sliuzas, RV 2018, Context-Based Filtering of Noisy Labels for Automatic Basemap Updating From UAV Data. in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing. 8 edn, vol. 11, IEEE Journal of selected topics in applied earth observations and remote sensing, pp. 1-11. https://doi.org/10.1109/JSTARS.2017.2762905

Context-Based Filtering of Noisy Labels for Automatic Basemap Updating From UAV Data. / Gevaert, C.M.; Persello, C.; Oude Elberink, S.J.; Vosselman, G.; Sliuzas, R.V.

IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing. Vol. 11 8. ed. 2018. p. 1-11 (IEEE Journal of selected topics in applied earth observations and remote sensing).

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

TY - GEN

T1 - Context-Based Filtering of Noisy Labels for Automatic Basemap Updating From UAV Data

AU - Gevaert, C.M.

AU - Persello, C.

AU - Oude Elberink, S.J.

AU - Vosselman, G.

AU - Sliuzas, R.V.

PY - 2018/8

Y1 - 2018/8

N2 - Unmanned aerial vehicles (UAVs) have the potential to obtain high-resolution aerial imagery at frequent intervals, making them a valuable tool for urban planners who require up-to-date basemaps. Supervised classification methods can be exploited to translate the UAV data into such basemaps. However, these methods require labeled training samples, the collection of which may be complex and time consuming. Existing spatial datasets can be exploited to provide the training labels, but these often contain errors due to differences in the date or resolution of the dataset from which these outdated labels were obtained. In this paper, we propose an approach for updating basemaps using global and local contextual cues to automatically remove unreliable samples from the training set, and thereby, improve the classification accuracy. Using UAV datasets over Kigali, Rwanda, and Dar es Salaam, Tanzania, we demonstrate how the amount of mislabeled training samples can be reduced by 44.1% and 35.5%, respectively, leading to a classification accuracy of 92.1% in Kigali and 91.3% in Dar es Salaam. To achieve the same accuracy in Dar es Salaam, between 50000 and 60000 manually labeled image segments would be needed. This demonstrates that the proposed approach of using outdated spatial data to provide labels and iteratively removing unreliable samples is a viable method for obtaining high classification accuracies while reducing the costly step of acquiring labeled training samples.

AB - Unmanned aerial vehicles (UAVs) have the potential to obtain high-resolution aerial imagery at frequent intervals, making them a valuable tool for urban planners who require up-to-date basemaps. Supervised classification methods can be exploited to translate the UAV data into such basemaps. However, these methods require labeled training samples, the collection of which may be complex and time consuming. Existing spatial datasets can be exploited to provide the training labels, but these often contain errors due to differences in the date or resolution of the dataset from which these outdated labels were obtained. In this paper, we propose an approach for updating basemaps using global and local contextual cues to automatically remove unreliable samples from the training set, and thereby, improve the classification accuracy. Using UAV datasets over Kigali, Rwanda, and Dar es Salaam, Tanzania, we demonstrate how the amount of mislabeled training samples can be reduced by 44.1% and 35.5%, respectively, leading to a classification accuracy of 92.1% in Kigali and 91.3% in Dar es Salaam. To achieve the same accuracy in Dar es Salaam, between 50000 and 60000 manually labeled image segments would be needed. This demonstrates that the proposed approach of using outdated spatial data to provide labels and iteratively removing unreliable samples is a viable method for obtaining high classification accuracies while reducing the costly step of acquiring labeled training samples.

KW - ITC-ISI-JOURNAL-ARTICLE

UR - https://ezproxy2.utwente.nl/login?url=https://webapps.itc.utwente.nl/library/2017/isi/persello_con.pdf

U2 - 10.1109/JSTARS.2017.2762905

DO - 10.1109/JSTARS.2017.2762905

M3 - Conference contribution

VL - 11

T3 - IEEE Journal of selected topics in applied earth observations and remote sensing

SP - 1

EP - 11

BT - IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing

ER -

Gevaert CM, Persello C, Oude Elberink SJ, Vosselman G, Sliuzas RV. Context-Based Filtering of Noisy Labels for Automatic Basemap Updating From UAV Data. In IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing. 8 ed. Vol. 11. 2018. p. 1-11. (IEEE Journal of selected topics in applied earth observations and remote sensing). https://doi.org/10.1109/JSTARS.2017.2762905