Auxiliary datasets improve accuracy of object-based land use/land cover classification in heterogeneous savanna landscapes

Tutkimustuotos: ArtikkelijulkaisuArtikkeliTieteellinenvertaisarvioitu

Kuvaus

Classifying land use/land cover (LULC) with sufficient accuracy in heterogeneous landscapes is challenging
using only satellite imagery. To improve classification accuracy inclusion of features from auxiliary geospatial
datasets in classification models is applied since 1980s. However, the method is mostly limited to pixel-based
classifications, and the coverage, accuracy and resolution of free and open-access auxiliary datasets have been
poor until recent years.
We evaluated how recent global coverage open-access geospatial datasets improve object-based LULC classification
accuracy compared to using only spectral and texture features from satellite images. We applied
feature sets topography, population, soil, canopy cover, distance to watercourses and spectral-temporal metrics
from Landsat-8 time series on the southern foothills and savanna of Mt. Kilimanjaro, Tanzania, where the
landscape is characterized by heterogeneous and fragmented mosaic of disturbed savanna vegetation, croplands,
and settlements.
The classification was based on image objects (groups of spectrally similar pixels) derived from segmentation
of four Formosat-2 scenes with 8m spatial resolution using 1370 ground reference points for training, validation,
and for defining 17 LULC classes. We built six Random Forest classification models with different sets of object
features in each. The baseline model having only spectral and texture features was compared with five other
models supplemented with auxiliary features.
Inclusion of auxiliary features significantly improved classification overall accuracy (OA). The baseline model
gave a median OA of 60.7%, but auxiliary features in other models increased median OA between 6.1 and 16.5
percentage points. The best OA was achieved with a model including all features of which elevation was the most
important auxiliary feature followed by Enhanced Vegetation Index temporal range and slope degree.
Applying object-based classification to geospatial information on topography, soil, settlement patterns and
vegetation phenology, the discriminatory potential of challenging LULC classes can be significantly improved.
We demonstrated this for the first time, and the technique shows good potential for improving LULC mapping
across a multitude of fragmented landscapes worldwide.
Alkuperäiskielienglanti
Artikkeli111354
LehtiRemote Sensing of Environment
Vuosikerta233
Numero111354
Sivumäärä17
ISSN0034-4257
DOI - pysyväislinkit
TilaJulkaistu - marraskuuta 2019
OKM-julkaisutyyppiA1 Alkuperäisartikkeli tieteellisessä aikakauslehdessä, vertaisarvioitu

Tieteenalat

  • 1171 Geotieteet
  • image segmentation
  • OBIA
  • Land use
  • Land cover classification
  • Auxiliary data
  • Random Forest
  • Satellite Time Series

Lainaa tätä

@article{9d4b9e8618f44c4ca564d14489c14d24,
title = "Auxiliary datasets improve accuracy of object-based land use/land cover classification in heterogeneous savanna landscapes",
abstract = "Classifying land use/land cover (LULC) with sufficient accuracy in heterogeneous landscapes is challengingusing only satellite imagery. To improve classification accuracy inclusion of features from auxiliary geospatialdatasets in classification models is applied since 1980s. However, the method is mostly limited to pixel-basedclassifications, and the coverage, accuracy and resolution of free and open-access auxiliary datasets have beenpoor until recent years.We evaluated how recent global coverage open-access geospatial datasets improve object-based LULC classificationaccuracy compared to using only spectral and texture features from satellite images. We appliedfeature sets topography, population, soil, canopy cover, distance to watercourses and spectral-temporal metricsfrom Landsat-8 time series on the southern foothills and savanna of Mt. Kilimanjaro, Tanzania, where thelandscape is characterized by heterogeneous and fragmented mosaic of disturbed savanna vegetation, croplands,and settlements.The classification was based on image objects (groups of spectrally similar pixels) derived from segmentationof four Formosat-2 scenes with 8m spatial resolution using 1370 ground reference points for training, validation,and for defining 17 LULC classes. We built six Random Forest classification models with different sets of objectfeatures in each. The baseline model having only spectral and texture features was compared with five othermodels supplemented with auxiliary features.Inclusion of auxiliary features significantly improved classification overall accuracy (OA). The baseline modelgave a median OA of 60.7{\%}, but auxiliary features in other models increased median OA between 6.1 and 16.5percentage points. The best OA was achieved with a model including all features of which elevation was the mostimportant auxiliary feature followed by Enhanced Vegetation Index temporal range and slope degree.Applying object-based classification to geospatial information on topography, soil, settlement patterns andvegetation phenology, the discriminatory potential of challenging LULC classes can be significantly improved.We demonstrated this for the first time, and the technique shows good potential for improving LULC mappingacross a multitude of fragmented landscapes worldwide.",
keywords = "1171 Geosciences, image segmentation, OBIA, Land use, Land cover classification, Auxiliary data, Random Forest, Satellite Time Series, Image segmentation, OBIA, Land use/land cover mapping, Auxiliary data, Random Forest, Satellite Time Series, IMAGE-ANALYSIS, RANDOM FOREST, MT. KILIMANJARO, TIME-SERIES, SURFACE TEMPERATURE, SOUTHERN SLOPES, ANCILLARY DATA, VEGETATION, MULTISOURCE, SELECTION",
author = "Pekka Hurskainen and Hari Adhikari and Mika Siljander and Petri Pellikka and Andreas Hemp",
year = "2019",
month = "11",
doi = "10.1016/j.rse.2019.111354",
language = "English",
volume = "233",
journal = "Remote Sensing of Environment",
issn = "0034-4257",
publisher = "EXCERPTA MEDICA INC-ELSEVIER SCIENCE INC",
number = "111354",

}

Auxiliary datasets improve accuracy of object-based land use/land cover classification in heterogeneous savanna landscapes. / Hurskainen, Pekka; Adhikari, Hari; Siljander, Mika; Pellikka, Petri; Hemp, Andreas.

julkaisussa: Remote Sensing of Environment, Vuosikerta 233, Nro 111354, 111354, 11.2019.

Tutkimustuotos: ArtikkelijulkaisuArtikkeliTieteellinenvertaisarvioitu

TY - JOUR

T1 - Auxiliary datasets improve accuracy of object-based land use/land cover classification in heterogeneous savanna landscapes

AU - Hurskainen, Pekka

AU - Adhikari, Hari

AU - Siljander, Mika

AU - Pellikka, Petri

AU - Hemp, Andreas

PY - 2019/11

Y1 - 2019/11

N2 - Classifying land use/land cover (LULC) with sufficient accuracy in heterogeneous landscapes is challengingusing only satellite imagery. To improve classification accuracy inclusion of features from auxiliary geospatialdatasets in classification models is applied since 1980s. However, the method is mostly limited to pixel-basedclassifications, and the coverage, accuracy and resolution of free and open-access auxiliary datasets have beenpoor until recent years.We evaluated how recent global coverage open-access geospatial datasets improve object-based LULC classificationaccuracy compared to using only spectral and texture features from satellite images. We appliedfeature sets topography, population, soil, canopy cover, distance to watercourses and spectral-temporal metricsfrom Landsat-8 time series on the southern foothills and savanna of Mt. Kilimanjaro, Tanzania, where thelandscape is characterized by heterogeneous and fragmented mosaic of disturbed savanna vegetation, croplands,and settlements.The classification was based on image objects (groups of spectrally similar pixels) derived from segmentationof four Formosat-2 scenes with 8m spatial resolution using 1370 ground reference points for training, validation,and for defining 17 LULC classes. We built six Random Forest classification models with different sets of objectfeatures in each. The baseline model having only spectral and texture features was compared with five othermodels supplemented with auxiliary features.Inclusion of auxiliary features significantly improved classification overall accuracy (OA). The baseline modelgave a median OA of 60.7%, but auxiliary features in other models increased median OA between 6.1 and 16.5percentage points. The best OA was achieved with a model including all features of which elevation was the mostimportant auxiliary feature followed by Enhanced Vegetation Index temporal range and slope degree.Applying object-based classification to geospatial information on topography, soil, settlement patterns andvegetation phenology, the discriminatory potential of challenging LULC classes can be significantly improved.We demonstrated this for the first time, and the technique shows good potential for improving LULC mappingacross a multitude of fragmented landscapes worldwide.

AB - Classifying land use/land cover (LULC) with sufficient accuracy in heterogeneous landscapes is challengingusing only satellite imagery. To improve classification accuracy inclusion of features from auxiliary geospatialdatasets in classification models is applied since 1980s. However, the method is mostly limited to pixel-basedclassifications, and the coverage, accuracy and resolution of free and open-access auxiliary datasets have beenpoor until recent years.We evaluated how recent global coverage open-access geospatial datasets improve object-based LULC classificationaccuracy compared to using only spectral and texture features from satellite images. We appliedfeature sets topography, population, soil, canopy cover, distance to watercourses and spectral-temporal metricsfrom Landsat-8 time series on the southern foothills and savanna of Mt. Kilimanjaro, Tanzania, where thelandscape is characterized by heterogeneous and fragmented mosaic of disturbed savanna vegetation, croplands,and settlements.The classification was based on image objects (groups of spectrally similar pixels) derived from segmentationof four Formosat-2 scenes with 8m spatial resolution using 1370 ground reference points for training, validation,and for defining 17 LULC classes. We built six Random Forest classification models with different sets of objectfeatures in each. The baseline model having only spectral and texture features was compared with five othermodels supplemented with auxiliary features.Inclusion of auxiliary features significantly improved classification overall accuracy (OA). The baseline modelgave a median OA of 60.7%, but auxiliary features in other models increased median OA between 6.1 and 16.5percentage points. The best OA was achieved with a model including all features of which elevation was the mostimportant auxiliary feature followed by Enhanced Vegetation Index temporal range and slope degree.Applying object-based classification to geospatial information on topography, soil, settlement patterns andvegetation phenology, the discriminatory potential of challenging LULC classes can be significantly improved.We demonstrated this for the first time, and the technique shows good potential for improving LULC mappingacross a multitude of fragmented landscapes worldwide.

KW - 1171 Geosciences

KW - image segmentation

KW - OBIA

KW - Land use

KW - Land cover classification

KW - Auxiliary data

KW - Random Forest

KW - Satellite Time Series

KW - Image segmentation

KW - OBIA

KW - Land use/land cover mapping

KW - Auxiliary data

KW - Random Forest

KW - Satellite Time Series

KW - IMAGE-ANALYSIS

KW - RANDOM FOREST

KW - MT. KILIMANJARO

KW - TIME-SERIES

KW - SURFACE TEMPERATURE

KW - SOUTHERN SLOPES

KW - ANCILLARY DATA

KW - VEGETATION

KW - MULTISOURCE

KW - SELECTION

U2 - 10.1016/j.rse.2019.111354

DO - 10.1016/j.rse.2019.111354

M3 - Article

VL - 233

JO - Remote Sensing of Environment

JF - Remote Sensing of Environment

SN - 0034-4257

IS - 111354

M1 - 111354

ER -