Towards Operative Forest Inventory by Extraction of Tree Level Information from VHR Satellite Images

Heikki Astola, Heikki Ahola, Kaj Andersson, Tuomas Häme, Jorma Kilpi, Matthieu Molinier, Yrjö Rauste, Jussi Rasinmäki

Research output: Chapter in Book/Report/Conference proceedingConference article in proceedingsScientific

Abstract

The work related to this paper is part of an on-going study called NewForest - Renewal of Forest Resource Mapping. In this study the methodologies developed for individual tree crown (ITC) recognition and crown width estimation will be combined with forest variable estimates that are produced using features calculated from segmented VHR satellite image. A field visit to Karttula, Eastern Finland, was conducted to collect the class and geo-location information for 1164 ground objects (900 trees and 264 non-tree objects). These data were used for the classifier model and feature selection and for species classification accuracy assessment. For testing the classifier ability to predict tree species proportions, an independent set of 178 forest field inventory plots was used. Seven classes were defined: pine, spruce, deciduous, shadow, open area, bare ground, green vegetation. A modified Local maximum (LM) filtering technique was used for individual tree crown (ITC) detection. The spectral signatures of an ITC were sampled with a radius of r=1.5 m around the ITC brightest pixel (feature set A). Also a set of 9 contextual features were extracted from circular neighbourhood (r=7.25 m) around the ITC (feature set B). A classifier model and feature selection was performed. A 5NN classifier provided the best overall performance in tree species classification in terms of classification accuracy and generalization. The overall classification accuracy for the seven classes was 73.8% with feature set A using 5NN classifier. With feature sets A and B combined the accuracy was 74.1%. The average RMS errors in species proportion prediction were 2.6% with feature set A and 2.5% with feature sets A and B combined.
Original languageEnglish
Title of host publicationProceedings of the ESA Living Planet Symposium, Bergen, Norway (ESA SP-686, December 2010)
Place of PublicationNoordwijk, The Netherlands
PublisherEuropean Space Agency ESA
ISBN (Print)978-92-9221-250-6
Publication statusPublished - 2010
MoE publication typeNot Eligible
EventESA Living Planet Symposium 2010 - Bergen, Norway
Duration: 28 Jun 20102 Jul 2010

Conference

ConferenceESA Living Planet Symposium 2010
CountryNorway
CityBergen
Period28/06/102/07/10

Fingerprint

forest inventory
accuracy assessment
satellite image
forest resource
pixel
methodology
vegetation
prediction

Keywords

  • forestry
  • tree species classification

Cite this

Astola, H., Ahola, H., Andersson, K., Häme, T., Kilpi, J., Molinier, M., ... Rasinmäki, J. (2010). Towards Operative Forest Inventory by Extraction of Tree Level Information from VHR Satellite Images. In Proceedings of the ESA Living Planet Symposium, Bergen, Norway (ESA SP-686, December 2010) Noordwijk, The Netherlands: European Space Agency ESA.
Astola, Heikki ; Ahola, Heikki ; Andersson, Kaj ; Häme, Tuomas ; Kilpi, Jorma ; Molinier, Matthieu ; Rauste, Yrjö ; Rasinmäki, Jussi. / Towards Operative Forest Inventory by Extraction of Tree Level Information from VHR Satellite Images. Proceedings of the ESA Living Planet Symposium, Bergen, Norway (ESA SP-686, December 2010). Noordwijk, The Netherlands : European Space Agency ESA, 2010.
@inproceedings{d137b16625a24fb9bd6be8a6c8e38629,
title = "Towards Operative Forest Inventory by Extraction of Tree Level Information from VHR Satellite Images",
abstract = "The work related to this paper is part of an on-going study called NewForest - Renewal of Forest Resource Mapping. In this study the methodologies developed for individual tree crown (ITC) recognition and crown width estimation will be combined with forest variable estimates that are produced using features calculated from segmented VHR satellite image. A field visit to Karttula, Eastern Finland, was conducted to collect the class and geo-location information for 1164 ground objects (900 trees and 264 non-tree objects). These data were used for the classifier model and feature selection and for species classification accuracy assessment. For testing the classifier ability to predict tree species proportions, an independent set of 178 forest field inventory plots was used. Seven classes were defined: pine, spruce, deciduous, shadow, open area, bare ground, green vegetation. A modified Local maximum (LM) filtering technique was used for individual tree crown (ITC) detection. The spectral signatures of an ITC were sampled with a radius of r=1.5 m around the ITC brightest pixel (feature set A). Also a set of 9 contextual features were extracted from circular neighbourhood (r=7.25 m) around the ITC (feature set B). A classifier model and feature selection was performed. A 5NN classifier provided the best overall performance in tree species classification in terms of classification accuracy and generalization. The overall classification accuracy for the seven classes was 73.8{\%} with feature set A using 5NN classifier. With feature sets A and B combined the accuracy was 74.1{\%}. The average RMS errors in species proportion prediction were 2.6{\%} with feature set A and 2.5{\%} with feature sets A and B combined.",
keywords = "forestry, tree species classification",
author = "Heikki Astola and Heikki Ahola and Kaj Andersson and Tuomas H{\"a}me and Jorma Kilpi and Matthieu Molinier and Yrj{\"o} Rauste and Jussi Rasinm{\"a}ki",
note = "Project code: 32655",
year = "2010",
language = "English",
isbn = "978-92-9221-250-6",
booktitle = "Proceedings of the ESA Living Planet Symposium, Bergen, Norway (ESA SP-686, December 2010)",
publisher = "European Space Agency ESA",
address = "France",

}

Astola, H, Ahola, H, Andersson, K, Häme, T, Kilpi, J, Molinier, M, Rauste, Y & Rasinmäki, J 2010, Towards Operative Forest Inventory by Extraction of Tree Level Information from VHR Satellite Images. in Proceedings of the ESA Living Planet Symposium, Bergen, Norway (ESA SP-686, December 2010). European Space Agency ESA, Noordwijk, The Netherlands, ESA Living Planet Symposium 2010, Bergen, Norway, 28/06/10.

Towards Operative Forest Inventory by Extraction of Tree Level Information from VHR Satellite Images. / Astola, Heikki; Ahola, Heikki; Andersson, Kaj; Häme, Tuomas; Kilpi, Jorma; Molinier, Matthieu; Rauste, Yrjö; Rasinmäki, Jussi.

Proceedings of the ESA Living Planet Symposium, Bergen, Norway (ESA SP-686, December 2010). Noordwijk, The Netherlands : European Space Agency ESA, 2010.

Research output: Chapter in Book/Report/Conference proceedingConference article in proceedingsScientific

TY - GEN

T1 - Towards Operative Forest Inventory by Extraction of Tree Level Information from VHR Satellite Images

AU - Astola, Heikki

AU - Ahola, Heikki

AU - Andersson, Kaj

AU - Häme, Tuomas

AU - Kilpi, Jorma

AU - Molinier, Matthieu

AU - Rauste, Yrjö

AU - Rasinmäki, Jussi

N1 - Project code: 32655

PY - 2010

Y1 - 2010

N2 - The work related to this paper is part of an on-going study called NewForest - Renewal of Forest Resource Mapping. In this study the methodologies developed for individual tree crown (ITC) recognition and crown width estimation will be combined with forest variable estimates that are produced using features calculated from segmented VHR satellite image. A field visit to Karttula, Eastern Finland, was conducted to collect the class and geo-location information for 1164 ground objects (900 trees and 264 non-tree objects). These data were used for the classifier model and feature selection and for species classification accuracy assessment. For testing the classifier ability to predict tree species proportions, an independent set of 178 forest field inventory plots was used. Seven classes were defined: pine, spruce, deciduous, shadow, open area, bare ground, green vegetation. A modified Local maximum (LM) filtering technique was used for individual tree crown (ITC) detection. The spectral signatures of an ITC were sampled with a radius of r=1.5 m around the ITC brightest pixel (feature set A). Also a set of 9 contextual features were extracted from circular neighbourhood (r=7.25 m) around the ITC (feature set B). A classifier model and feature selection was performed. A 5NN classifier provided the best overall performance in tree species classification in terms of classification accuracy and generalization. The overall classification accuracy for the seven classes was 73.8% with feature set A using 5NN classifier. With feature sets A and B combined the accuracy was 74.1%. The average RMS errors in species proportion prediction were 2.6% with feature set A and 2.5% with feature sets A and B combined.

AB - The work related to this paper is part of an on-going study called NewForest - Renewal of Forest Resource Mapping. In this study the methodologies developed for individual tree crown (ITC) recognition and crown width estimation will be combined with forest variable estimates that are produced using features calculated from segmented VHR satellite image. A field visit to Karttula, Eastern Finland, was conducted to collect the class and geo-location information for 1164 ground objects (900 trees and 264 non-tree objects). These data were used for the classifier model and feature selection and for species classification accuracy assessment. For testing the classifier ability to predict tree species proportions, an independent set of 178 forest field inventory plots was used. Seven classes were defined: pine, spruce, deciduous, shadow, open area, bare ground, green vegetation. A modified Local maximum (LM) filtering technique was used for individual tree crown (ITC) detection. The spectral signatures of an ITC were sampled with a radius of r=1.5 m around the ITC brightest pixel (feature set A). Also a set of 9 contextual features were extracted from circular neighbourhood (r=7.25 m) around the ITC (feature set B). A classifier model and feature selection was performed. A 5NN classifier provided the best overall performance in tree species classification in terms of classification accuracy and generalization. The overall classification accuracy for the seven classes was 73.8% with feature set A using 5NN classifier. With feature sets A and B combined the accuracy was 74.1%. The average RMS errors in species proportion prediction were 2.6% with feature set A and 2.5% with feature sets A and B combined.

KW - forestry

KW - tree species classification

M3 - Conference article in proceedings

SN - 978-92-9221-250-6

BT - Proceedings of the ESA Living Planet Symposium, Bergen, Norway (ESA SP-686, December 2010)

PB - European Space Agency ESA

CY - Noordwijk, The Netherlands

ER -

Astola H, Ahola H, Andersson K, Häme T, Kilpi J, Molinier M et al. Towards Operative Forest Inventory by Extraction of Tree Level Information from VHR Satellite Images. In Proceedings of the ESA Living Planet Symposium, Bergen, Norway (ESA SP-686, December 2010). Noordwijk, The Netherlands: European Space Agency ESA. 2010