New iterative learning strategy to improve classification systems by using outlier detection techniques

C. Pelletier; S. Valero; J. Inglada; G. Dedieu; N. Champion

doi:10.1109/IGARSS.2017.8127796

New iterative learning strategy to improve classification systems by using outlier detection techniques

C. Pelletier
, S. Valero
, J. Inglada
, G. Dedieu
, N. Champion

Research output: Chapter in Book/Report/Conference proceeding › Conference Paper › Research

3 Citations (Scopus)

Abstract

The supervised classification of satellite image time series allows obtaining reliable land cover maps over large areas. However, their quality depends on the reference datasets used for training the classifier. In remote sensing, reference data may lack of timeliness and accuracy which leads to the presence of mislabeled data degrading the classification performances. This work presents an iterative learning framework to deal with noisy instances, that can be seen as outliers. Several outlier detection strategies, based on the well-known Random Forests (RF) ensemble classifier, are proposed, evaluated quantitatively, and then compared with traditional methods. Experimental results have been carried out by using synthetic and real datasets representing annual vegetation profiles.

Original language	English
Title of host publication	2017 IEEE International Geoscience & Remote Sensing Symposium - Proceedings
Subtitle of host publication	July 23–28, 2017 Fort Worth, Texas, USA
Editors	Joel T. Johnson, Kun-Shan Chen
Place of Publication	Piscataway NJ USA
Publisher	IEEE, Institute of Electrical and Electronics Engineers
Pages	3676-3679
Number of pages	4
ISBN (Electronic)	9781509049516
ISBN (Print)	9781509049523
DOIs	https://doi.org/10.1109/IGARSS.2017.8127796
Publication status	Published - 2017
Externally published	Yes
Event	IEEE International Geoscience and Remote Sensing Symposium 2017 - Fort Worth, United States of America Duration: 23 Jul 2017 → 28 Jul 2017 Conference number: 37th https://www2.securecms.com/IGARSS2017/Default.asp https://ieeexplore.ieee.org/xpl/conhome/8118204/proceeding (Proceedings)

Conference

Conference	IEEE International Geoscience and Remote Sensing Symposium 2017
Abbreviated title	IGARSS 2017
Country/Territory	United States of America
City	Fort Worth
Period	23/07/17 → 28/07/17
Internet address	https://www2.securecms.com/IGARSS2017/Default.asp https://ieeexplore.ieee.org/xpl/conhome/8118204/proceeding (Proceedings)

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 15 Life on Land

Keywords

Land cover mapping
Mislabeled data
Outlier detection
Random Forests
Satellite Image Time Series classification

Access to Document

10.1109/IGARSS.2017.8127796

Cite this

Pelletier, C., Valero, S., Inglada, J., Dedieu, G., & Champion, N. (2017). New iterative learning strategy to improve classification systems by using outlier detection techniques. In J. T. Johnson, & K.-S. Chen (Eds.), 2017 IEEE International Geoscience & Remote Sensing Symposium - Proceedings: July 23–28, 2017 Fort Worth, Texas, USA (pp. 3676-3679). IEEE, Institute of Electrical and Electronics Engineers. https://doi.org/10.1109/IGARSS.2017.8127796

Pelletier, C. ; Valero, S. ; Inglada, J. et al. / New iterative learning strategy to improve classification systems by using outlier detection techniques. 2017 IEEE International Geoscience & Remote Sensing Symposium - Proceedings: July 23–28, 2017 Fort Worth, Texas, USA. editor / Joel T. Johnson ; Kun-Shan Chen. Piscataway NJ USA : IEEE, Institute of Electrical and Electronics Engineers, 2017. pp. 3676-3679

@inproceedings{7d0dabdae0754a0bacca6d2ec5ccbc1e,

title = "New iterative learning strategy to improve classification systems by using outlier detection techniques",

abstract = "The supervised classification of satellite image time series allows obtaining reliable land cover maps over large areas. However, their quality depends on the reference datasets used for training the classifier. In remote sensing, reference data may lack of timeliness and accuracy which leads to the presence of mislabeled data degrading the classification performances. This work presents an iterative learning framework to deal with noisy instances, that can be seen as outliers. Several outlier detection strategies, based on the well-known Random Forests (RF) ensemble classifier, are proposed, evaluated quantitatively, and then compared with traditional methods. Experimental results have been carried out by using synthetic and real datasets representing annual vegetation profiles.",

keywords = "Land cover mapping, Mislabeled data, Outlier detection, Random Forests, Satellite Image Time Series classification",

author = "C. Pelletier and S. Valero and J. Inglada and G. Dedieu and N. Champion",

year = "2017",

doi = "10.1109/IGARSS.2017.8127796",

language = "English",

isbn = "9781509049523",

pages = "3676--3679",

editor = "\{T. Johnson\}, \{Joel \} and Chen, \{Kun-Shan \}",

booktitle = "2017 IEEE International Geoscience \& Remote Sensing Symposium - Proceedings",

publisher = "IEEE, Institute of Electrical and Electronics Engineers",

address = "United States of America",

note = "IEEE International Geoscience and Remote Sensing Symposium 2017, IGARSS 2017 ; Conference date: 23-07-2017 Through 28-07-2017",

url = "https://www2.securecms.com/IGARSS2017/Default.asp, https://ieeexplore.ieee.org/xpl/conhome/8118204/proceeding",

}

Pelletier, C, Valero, S, Inglada, J, Dedieu, G & Champion, N 2017, New iterative learning strategy to improve classification systems by using outlier detection techniques. in J T. Johnson & K-S Chen (eds), 2017 IEEE International Geoscience & Remote Sensing Symposium - Proceedings: July 23–28, 2017 Fort Worth, Texas, USA. IEEE, Institute of Electrical and Electronics Engineers, Piscataway NJ USA, pp. 3676-3679, IEEE International Geoscience and Remote Sensing Symposium 2017, Fort Worth, Texas, United States of America, 23/07/17. https://doi.org/10.1109/IGARSS.2017.8127796

New iterative learning strategy to improve classification systems by using outlier detection techniques. / Pelletier, C.; Valero, S.; Inglada, J. et al.
2017 IEEE International Geoscience & Remote Sensing Symposium - Proceedings: July 23–28, 2017 Fort Worth, Texas, USA. ed. / Joel T. Johnson; Kun-Shan Chen. Piscataway NJ USA: IEEE, Institute of Electrical and Electronics Engineers, 2017. p. 3676-3679.

Research output: Chapter in Book/Report/Conference proceeding › Conference Paper › Research

TY - GEN

T1 - New iterative learning strategy to improve classification systems by using outlier detection techniques

AU - Pelletier, C.

AU - Valero, S.

AU - Inglada, J.

AU - Dedieu, G.

AU - Champion, N.

N1 - Conference code: 37th

PY - 2017

Y1 - 2017

N2 - The supervised classification of satellite image time series allows obtaining reliable land cover maps over large areas. However, their quality depends on the reference datasets used for training the classifier. In remote sensing, reference data may lack of timeliness and accuracy which leads to the presence of mislabeled data degrading the classification performances. This work presents an iterative learning framework to deal with noisy instances, that can be seen as outliers. Several outlier detection strategies, based on the well-known Random Forests (RF) ensemble classifier, are proposed, evaluated quantitatively, and then compared with traditional methods. Experimental results have been carried out by using synthetic and real datasets representing annual vegetation profiles.

AB - The supervised classification of satellite image time series allows obtaining reliable land cover maps over large areas. However, their quality depends on the reference datasets used for training the classifier. In remote sensing, reference data may lack of timeliness and accuracy which leads to the presence of mislabeled data degrading the classification performances. This work presents an iterative learning framework to deal with noisy instances, that can be seen as outliers. Several outlier detection strategies, based on the well-known Random Forests (RF) ensemble classifier, are proposed, evaluated quantitatively, and then compared with traditional methods. Experimental results have been carried out by using synthetic and real datasets representing annual vegetation profiles.

KW - Land cover mapping

KW - Mislabeled data

KW - Outlier detection

KW - Random Forests

KW - Satellite Image Time Series classification

UR - https://www.scopus.com/pages/publications/85041842537

U2 - 10.1109/IGARSS.2017.8127796

DO - 10.1109/IGARSS.2017.8127796

M3 - Conference Paper

AN - SCOPUS:85041842537

SN - 9781509049523

SP - 3676

EP - 3679

BT - 2017 IEEE International Geoscience & Remote Sensing Symposium - Proceedings

A2 - T. Johnson, Joel

A2 - Chen, Kun-Shan

PB - IEEE, Institute of Electrical and Electronics Engineers

CY - Piscataway NJ USA

T2 - IEEE International Geoscience and Remote Sensing Symposium 2017

Y2 - 23 July 2017 through 28 July 2017

ER -

Pelletier C, Valero S, Inglada J, Dedieu G, Champion N. New iterative learning strategy to improve classification systems by using outlier detection techniques. In T. Johnson J, Chen KS, editors, 2017 IEEE International Geoscience & Remote Sensing Symposium - Proceedings: July 23–28, 2017 Fort Worth, Texas, USA. Piscataway NJ USA: IEEE, Institute of Electrical and Electronics Engineers. 2017. p. 3676-3679 doi: 10.1109/IGARSS.2017.8127796

New iterative learning strategy to improve classification systems by using outlier detection techniques

Abstract

Conference

UN SDGs

Keywords

Access to Document

Other files and links

Cite this