Tree species classification using multi-temporal Sentinel-2 data
The Sentinel-2 program provides the opportunity to monitor terrestrial ecosystems with a high temporal- and spectral resolution. In this study, the utilization of multi-temporal Sentinel-2 imagery and it’s spectral variation due to phenology for classification of common tree species is evaluated at...
| Autor principal: | |
|---|---|
| Formato: | H3 |
| Lenguaje: | Inglés sueco |
| Publicado: |
SLU/Dept. of Forest Resource Management
2018
|
| Materias: |
| _version_ | 1855572294537052160 |
|---|---|
| author | Persson, Magnus |
| author_browse | Persson, Magnus |
| author_facet | Persson, Magnus |
| author_sort | Persson, Magnus |
| collection | Epsilon Archive for Student Projects |
| description | The Sentinel-2 program provides the opportunity to monitor terrestrial ecosystems with a high temporal- and spectral resolution. In this study, the utilization of multi-temporal Sentinel-2 imagery and it’s spectral variation due to phenology for classification of common tree species is evaluated at the forest estate Remningstorp in central Sweden.
The tree species classes to be classified were: Norway Spruce (Picea abies), Scots Pine (Pinus silvestris), Hybrid Larch (Larix × marschlinsii), Silver Birch (Betula pendula) and Pedunculate Oak (Quercus rubur). The Random Forest classifier (RF) was fitted to four Sentinel-2 images taken during the vegetation period of 2017. The RF classifier was also coupled with the feature selection algorithm Recursive Feature Elimination to form a model with an optimal subset of bands. In addition to the classification, spectral profile plots were constructed for each species to visualize the possibility for identifying the less represented tree species.
The use of four satellite images from April 7th, May 27th, July 9th and October 19th resulted in a higher overall accuracy (86.4 %) compared to using single images (71.5 % – 79.4 %). The late spring image (May 27th) was found to be important since it always was included in the most accurate classifications, independently of the number of images.
The best combination of bands resulted in a model with 87.6 % in overall accuracy and included 37 of 40 bands. The highest ranked bands were all May bands except the red band, the SWIR 1-2 and red bands from April, July and October. The 5 tree species classes were classified with satisfying results and the Producer’s Accuracy ranged from 73.7 % to 97.4 %. |
| format | H3 |
| id | RepoSLU13740 |
| institution | Swedish University of Agricultural Sciences |
| language | Inglés swe |
| publishDate | 2018 |
| publishDateSort | 2018 |
| publisher | SLU/Dept. of Forest Resource Management |
| publisherStr | SLU/Dept. of Forest Resource Management |
| record_format | eprints |
| spelling | RepoSLU137402019-02-27T11:51:32Z Tree species classification using multi-temporal Sentinel-2 data Trädslagsklassificering med multi-temporalt Sentinel-2 data Persson, Magnus sentinel-2 tree species random forest recursive feature elimination The Sentinel-2 program provides the opportunity to monitor terrestrial ecosystems with a high temporal- and spectral resolution. In this study, the utilization of multi-temporal Sentinel-2 imagery and it’s spectral variation due to phenology for classification of common tree species is evaluated at the forest estate Remningstorp in central Sweden. The tree species classes to be classified were: Norway Spruce (Picea abies), Scots Pine (Pinus silvestris), Hybrid Larch (Larix × marschlinsii), Silver Birch (Betula pendula) and Pedunculate Oak (Quercus rubur). The Random Forest classifier (RF) was fitted to four Sentinel-2 images taken during the vegetation period of 2017. The RF classifier was also coupled with the feature selection algorithm Recursive Feature Elimination to form a model with an optimal subset of bands. In addition to the classification, spectral profile plots were constructed for each species to visualize the possibility for identifying the less represented tree species. The use of four satellite images from April 7th, May 27th, July 9th and October 19th resulted in a higher overall accuracy (86.4 %) compared to using single images (71.5 % – 79.4 %). The late spring image (May 27th) was found to be important since it always was included in the most accurate classifications, independently of the number of images. The best combination of bands resulted in a model with 87.6 % in overall accuracy and included 37 of 40 bands. The highest ranked bands were all May bands except the red band, the SWIR 1-2 and red bands from April, July and October. The 5 tree species classes were classified with satisfying results and the Producer’s Accuracy ranged from 73.7 % to 97.4 %. Sentinel-2 satelliterna möjliggör övervakning av terrestra ekosystem med hög temporal- och spektral upplösning. I denna studie utvärderas möjligheten att nyttja den fenologiska variationen för att klassificera Sveriges vanligaste trädslag på skogsfastigheten Remningstorp i Västra Götaland. I denna studie användes ett multi-temporalt Sentinel-2 dataset för att klassificera gran (Picea abies), tall (Pinus silvestris), hybridlärk (Larix × marschlinsii), vårtbjörk (Betula pendula) och skogsek (Quercus rubur). Klassificeringsmetoden Random Forest (RF) användes för att utvärdera prestandan för olika kombinationer av fyra satellitbilder spridda över vegetationsperioden 2017. Recursive Feature Elimination (RFE) användes också tillsammans med RF för att hitta ett urval av band som bidrog mest till klassificeringens noggrannhet. Dessutom skapades spektrala kurvor för alla trädslag som komplement till klassificeringen och för att visualisera möjligheten att urskilja de mindre förkommande trädslagen på studieområdet. Det multi-temporala datasetet innehållande alla satellitbilder (7 april, 27 maj, 9 juli och 19 oktober) resulterade i en noggrannhet på 86.4 %. Det är avsevärt bättre resultat än att endast använda enskilda satellitbilder (71.5 % – 79.4 %). Maj bilden var viktig då den alltid var med i den bästa modellen, oberoende av vilken av de andra satellitbilderna den kombinerades med. Den bästa modellen från RFE resulterade i 87.6 % noggrannhet och innehöll 37 av 40 band. Enligt rankingen från RFE-modellen var de viktigaste banden alla band från maj-bilden utom det röda bandet samt SWIR 1–2 banden från april-, juli- och oktoberbilderna. Dessa resultat stärks av de spektrala kurvorna. Höga värden för ”Producer’s Accuracy” erhölls för gran, skogsek och vårtbjörk (90 %, 97.4 %, 95.6%), medan medelgoda värden erhölls för hybridlärk och tall (81.5 %, 73.7 %). SLU/Dept. of Forest Resource Management 2018 H3 eng swe https://stud.epsilon.slu.se/13740/ |
| spellingShingle | sentinel-2 tree species random forest recursive feature elimination Persson, Magnus Tree species classification using multi-temporal Sentinel-2 data |
| title | Tree species classification using multi-temporal Sentinel-2 data |
| title_full | Tree species classification using multi-temporal Sentinel-2 data |
| title_fullStr | Tree species classification using multi-temporal Sentinel-2 data |
| title_full_unstemmed | Tree species classification using multi-temporal Sentinel-2 data |
| title_short | Tree species classification using multi-temporal Sentinel-2 data |
| title_sort | tree species classification using multi-temporal sentinel-2 data |
| topic | sentinel-2 tree species random forest recursive feature elimination |