Feature Relevance Assessment Of Multispectral Airborne Lidar Data For Tree Species Classification

N. Amiri, M. Heurich, P. Krzystek, A. K. Skidmore

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

2 Citations (Scopus)
46 Downloads (Pure)

Abstract

The presented experiment investigates the potential of Multispectral Laser Scanning (MLS) point clouds for single tree species classification. The basic idea is to simulate a MLS sensor by combining two different Lidar sensors providing three different wavel-ngthes. The available data were acquired in the summer 2016 at the same date in a leaf-on condition with an average point density of 37 points/m 2. For the purpose of classification, we segmented the combined 3D point clouds consisiting of three different spectral channels into 3D clusters using Normalized Cut segmentation approach. Then, we extracted four group of features from the 3D point cloud space. Once a varity of features has been extracted, we applied forward stepwise feature selection in order to reduce the number of irrelevant or redundant features. For the classification, we used multinomial logestic regression with L 1 regularization. Our study is conducted using 586 ground measured single trees from 20 sample plots in the Bavarian Forest National Park, in Germany. Due to lack of reference data for some rare species, we focused on four classes of species. The results show an improvement between 4-10 pp for the tree species classification by using MLS data in comparison to a single wavelength based approach. A cross validated (15-fold) accuracy of 0.75 can be achieved when all feature sets from three different spectral channels are used. Our results cleary indicates that the use of MLS point clouds has great potential to improve detailed forest species mapping.

Original languageEnglish
Title of host publicationInternational Archives of Photogrammetry and Remote Sensing, ISPRS TC III Mid-term symposium
Subtitle of host publicationDevelopments, technologies and applications in Remote Sensing, 7-10 May, Beijing, China
EditorsJ. Jiang, A. Shaker, H. Zhang
PublisherInternational Society for Photogrammetry and Remote Sensing (ISPRS)
Pages31-34
Number of pages4
DOIs
Publication statusPublished - 30 Apr 2018
EventDevelopments, Technologies and Applications in Remote Sensing 2018: ISPRS TC III Symposium - Beijing International Convention Center, Beijing, China
Duration: 7 May 201810 May 2018
http://www.isprs-tc3.tianditu.com/

Publication series

NameInternational Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences - ISPRS Archives
PublisherCopernicus
VolumeXLII-3
ISSN (Print)1682-1750

Conference

ConferenceDevelopments, Technologies and Applications in Remote Sensing 2018
CountryChina
CityBeijing
Period7/05/1810/05/18
Internet address

Fingerprint

lidar
laser
sensor
rare species
segmentation
national park
fold
wavelength
summer
experiment

Keywords

  • ITC-GOLD

Cite this

Amiri, N., Heurich, M., Krzystek, P., & Skidmore, A. K. (2018). Feature Relevance Assessment Of Multispectral Airborne Lidar Data For Tree Species Classification. In J. Jiang, A. Shaker, & H. Zhang (Eds.), International Archives of Photogrammetry and Remote Sensing, ISPRS TC III Mid-term symposium : Developments, technologies and applications in Remote Sensing, 7-10 May, Beijing, China (pp. 31-34). (International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences - ISPRS Archives; Vol. XLII-3). International Society for Photogrammetry and Remote Sensing (ISPRS). https://doi.org/10.5194/isprs-archives-XLII-3-31-2018
Amiri, N. ; Heurich, M. ; Krzystek, P. ; Skidmore, A. K. / Feature Relevance Assessment Of Multispectral Airborne Lidar Data For Tree Species Classification. International Archives of Photogrammetry and Remote Sensing, ISPRS TC III Mid-term symposium : Developments, technologies and applications in Remote Sensing, 7-10 May, Beijing, China. editor / J. Jiang ; A. Shaker ; H. Zhang. International Society for Photogrammetry and Remote Sensing (ISPRS), 2018. pp. 31-34 (International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences - ISPRS Archives).
@inproceedings{f64faa158b1c445fb88ff5ef39752240,
title = "Feature Relevance Assessment Of Multispectral Airborne Lidar Data For Tree Species Classification",
abstract = "The presented experiment investigates the potential of Multispectral Laser Scanning (MLS) point clouds for single tree species classification. The basic idea is to simulate a MLS sensor by combining two different Lidar sensors providing three different wavel-ngthes. The available data were acquired in the summer 2016 at the same date in a leaf-on condition with an average point density of 37 points/m 2. For the purpose of classification, we segmented the combined 3D point clouds consisiting of three different spectral channels into 3D clusters using Normalized Cut segmentation approach. Then, we extracted four group of features from the 3D point cloud space. Once a varity of features has been extracted, we applied forward stepwise feature selection in order to reduce the number of irrelevant or redundant features. For the classification, we used multinomial logestic regression with L 1 regularization. Our study is conducted using 586 ground measured single trees from 20 sample plots in the Bavarian Forest National Park, in Germany. Due to lack of reference data for some rare species, we focused on four classes of species. The results show an improvement between 4-10 pp for the tree species classification by using MLS data in comparison to a single wavelength based approach. A cross validated (15-fold) accuracy of 0.75 can be achieved when all feature sets from three different spectral channels are used. Our results cleary indicates that the use of MLS point clouds has great potential to improve detailed forest species mapping.",
keywords = "ITC-GOLD",
author = "N. Amiri and M. Heurich and P. Krzystek and Skidmore, {A. K.}",
year = "2018",
month = "4",
day = "30",
doi = "10.5194/isprs-archives-XLII-3-31-2018",
language = "English",
series = "International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences - ISPRS Archives",
publisher = "International Society for Photogrammetry and Remote Sensing (ISPRS)",
pages = "31--34",
editor = "J. Jiang and A. Shaker and H. Zhang",
booktitle = "International Archives of Photogrammetry and Remote Sensing, ISPRS TC III Mid-term symposium",

}

Amiri, N, Heurich, M, Krzystek, P & Skidmore, AK 2018, Feature Relevance Assessment Of Multispectral Airborne Lidar Data For Tree Species Classification. in J Jiang, A Shaker & H Zhang (eds), International Archives of Photogrammetry and Remote Sensing, ISPRS TC III Mid-term symposium : Developments, technologies and applications in Remote Sensing, 7-10 May, Beijing, China. International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences - ISPRS Archives, vol. XLII-3, International Society for Photogrammetry and Remote Sensing (ISPRS), pp. 31-34, Developments, Technologies and Applications in Remote Sensing 2018, Beijing, China, 7/05/18. https://doi.org/10.5194/isprs-archives-XLII-3-31-2018

Feature Relevance Assessment Of Multispectral Airborne Lidar Data For Tree Species Classification. / Amiri, N.; Heurich, M.; Krzystek, P.; Skidmore, A. K.

International Archives of Photogrammetry and Remote Sensing, ISPRS TC III Mid-term symposium : Developments, technologies and applications in Remote Sensing, 7-10 May, Beijing, China. ed. / J. Jiang; A. Shaker; H. Zhang. International Society for Photogrammetry and Remote Sensing (ISPRS), 2018. p. 31-34 (International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences - ISPRS Archives; Vol. XLII-3).

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

TY - GEN

T1 - Feature Relevance Assessment Of Multispectral Airborne Lidar Data For Tree Species Classification

AU - Amiri, N.

AU - Heurich, M.

AU - Krzystek, P.

AU - Skidmore, A. K.

PY - 2018/4/30

Y1 - 2018/4/30

N2 - The presented experiment investigates the potential of Multispectral Laser Scanning (MLS) point clouds for single tree species classification. The basic idea is to simulate a MLS sensor by combining two different Lidar sensors providing three different wavel-ngthes. The available data were acquired in the summer 2016 at the same date in a leaf-on condition with an average point density of 37 points/m 2. For the purpose of classification, we segmented the combined 3D point clouds consisiting of three different spectral channels into 3D clusters using Normalized Cut segmentation approach. Then, we extracted four group of features from the 3D point cloud space. Once a varity of features has been extracted, we applied forward stepwise feature selection in order to reduce the number of irrelevant or redundant features. For the classification, we used multinomial logestic regression with L 1 regularization. Our study is conducted using 586 ground measured single trees from 20 sample plots in the Bavarian Forest National Park, in Germany. Due to lack of reference data for some rare species, we focused on four classes of species. The results show an improvement between 4-10 pp for the tree species classification by using MLS data in comparison to a single wavelength based approach. A cross validated (15-fold) accuracy of 0.75 can be achieved when all feature sets from three different spectral channels are used. Our results cleary indicates that the use of MLS point clouds has great potential to improve detailed forest species mapping.

AB - The presented experiment investigates the potential of Multispectral Laser Scanning (MLS) point clouds for single tree species classification. The basic idea is to simulate a MLS sensor by combining two different Lidar sensors providing three different wavel-ngthes. The available data were acquired in the summer 2016 at the same date in a leaf-on condition with an average point density of 37 points/m 2. For the purpose of classification, we segmented the combined 3D point clouds consisiting of three different spectral channels into 3D clusters using Normalized Cut segmentation approach. Then, we extracted four group of features from the 3D point cloud space. Once a varity of features has been extracted, we applied forward stepwise feature selection in order to reduce the number of irrelevant or redundant features. For the classification, we used multinomial logestic regression with L 1 regularization. Our study is conducted using 586 ground measured single trees from 20 sample plots in the Bavarian Forest National Park, in Germany. Due to lack of reference data for some rare species, we focused on four classes of species. The results show an improvement between 4-10 pp for the tree species classification by using MLS data in comparison to a single wavelength based approach. A cross validated (15-fold) accuracy of 0.75 can be achieved when all feature sets from three different spectral channels are used. Our results cleary indicates that the use of MLS point clouds has great potential to improve detailed forest species mapping.

KW - ITC-GOLD

UR - https://ezproxy2.utwente.nl/login?url=https://webapps.itc.utwente.nl/library/2018/conf/skidmore_fea.pdf

U2 - 10.5194/isprs-archives-XLII-3-31-2018

DO - 10.5194/isprs-archives-XLII-3-31-2018

M3 - Conference contribution

T3 - International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences - ISPRS Archives

SP - 31

EP - 34

BT - International Archives of Photogrammetry and Remote Sensing, ISPRS TC III Mid-term symposium

A2 - Jiang, J.

A2 - Shaker, A.

A2 - Zhang, H.

PB - International Society for Photogrammetry and Remote Sensing (ISPRS)

ER -

Amiri N, Heurich M, Krzystek P, Skidmore AK. Feature Relevance Assessment Of Multispectral Airborne Lidar Data For Tree Species Classification. In Jiang J, Shaker A, Zhang H, editors, International Archives of Photogrammetry and Remote Sensing, ISPRS TC III Mid-term symposium : Developments, technologies and applications in Remote Sensing, 7-10 May, Beijing, China. International Society for Photogrammetry and Remote Sensing (ISPRS). 2018. p. 31-34. (International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences - ISPRS Archives). https://doi.org/10.5194/isprs-archives-XLII-3-31-2018