Evaluated density estimates of young forest stands using high resolution 2D imagery from UAV
After more than three decades of advanced technology and methods to evaluate young forest attributes, the forestry still prefers laborious field surveys in regenera-tion forests. Preceding studies have virtually all successfully identified the saplings, but the cost has outweighed the benefits or at...
| Autor principal: | |
|---|---|
| Formato: | Second cycle, A2E |
| Lenguaje: | sueco Inglés |
| Publicado: |
2020
|
| Materias: | |
| Acceso en línea: | https://stud.epsilon.slu.se/15343/ |
| _version_ | 1855572562390548480 |
|---|---|
| author | Kingstad, Victor |
| author_browse | Kingstad, Victor |
| author_facet | Kingstad, Victor |
| author_sort | Kingstad, Victor |
| collection | Epsilon Archive for Student Projects |
| description | After more than three decades of advanced technology and methods to evaluate young forest attributes, the forestry still prefers laborious field surveys in regenera-tion forests. Preceding studies have virtually all successfully identified the saplings, but the cost has outweighed the benefits or at least not yet convinced the companies to step away from the traditional field surveys; high-resolution photogrammetric point clouds and ALS-data can provide accurate estimations of the biophysical prop-erties but requires intense data processing. One way to reduce the costs could be to use less accurate 2D imagery by scaling images to the approximate altitude.
The fundamentals of using cheap low-tech UAV imagery to aid management of young boreal forests was explored in this study, the results were obtained by com-paring image interpreted saplings from images processed by Dianthus Rapid Drone Map™ to Sveaskog’s traditional inventory practices of young forest stands. Three main issues were evaluated: ‘Camera positioning and image scale’, ‘Sapling classifi-cation’ and ‘Need for cleaning assessment’. The study area included 57 forest stands and 290 sample plots within the county of Västerbotten, northern Sweden. The ma-terial & methods in the study were mainly predetermined by available equipment and field instructions from Sveaskog to enable efficient data collection through combined labour- and research work.
The results revealed that manual interpretation of the acquired images could dis-criminate between forest stands in need of cleaning and not with 82% accuracy, though the individual sapling counts comprised large errors, (Overall RMSE 6568 stems ha-1). Possible advantages compared to traditional surveys are the production of up to date high-quality maps for the brush-cutter operators, improved planning of retention delineation and most of all reduced time spent in field since over 1 million hectares are estimated in need of cleaning in Sweden annually.
This study indicated that simple UAV imagery without proper photogrammetry could be an alternative; the quality might be bad but possibly sufficient. |
| format | Second cycle, A2E |
| id | RepoSLU15343 |
| institution | Swedish University of Agricultural Sciences |
| language | Swedish Inglés |
| publishDate | 2020 |
| publishDateSort | 2020 |
| record_format | eprints |
| spelling | RepoSLU153432020-02-21T02:00:48Z https://stud.epsilon.slu.se/15343/ Evaluated density estimates of young forest stands using high resolution 2D imagery from UAV Kingstad, Victor Forestry - General aspects Forestry production After more than three decades of advanced technology and methods to evaluate young forest attributes, the forestry still prefers laborious field surveys in regenera-tion forests. Preceding studies have virtually all successfully identified the saplings, but the cost has outweighed the benefits or at least not yet convinced the companies to step away from the traditional field surveys; high-resolution photogrammetric point clouds and ALS-data can provide accurate estimations of the biophysical prop-erties but requires intense data processing. One way to reduce the costs could be to use less accurate 2D imagery by scaling images to the approximate altitude. The fundamentals of using cheap low-tech UAV imagery to aid management of young boreal forests was explored in this study, the results were obtained by com-paring image interpreted saplings from images processed by Dianthus Rapid Drone Map™ to Sveaskog’s traditional inventory practices of young forest stands. Three main issues were evaluated: ‘Camera positioning and image scale’, ‘Sapling classifi-cation’ and ‘Need for cleaning assessment’. The study area included 57 forest stands and 290 sample plots within the county of Västerbotten, northern Sweden. The ma-terial & methods in the study were mainly predetermined by available equipment and field instructions from Sveaskog to enable efficient data collection through combined labour- and research work. The results revealed that manual interpretation of the acquired images could dis-criminate between forest stands in need of cleaning and not with 82% accuracy, though the individual sapling counts comprised large errors, (Overall RMSE 6568 stems ha-1). Possible advantages compared to traditional surveys are the production of up to date high-quality maps for the brush-cutter operators, improved planning of retention delineation and most of all reduced time spent in field since over 1 million hectares are estimated in need of cleaning in Sweden annually. This study indicated that simple UAV imagery without proper photogrammetry could be an alternative; the quality might be bad but possibly sufficient. Efter mer än tre decennier av utveckling av ny teknik och avancerade metoder för att inventera ungskogar med hög precision, föredrar skogsbruket fortfarande att genom-föra tidskrävande fältundersökningar för ändamålet. Tidigare fjärranlysstudier har identifierat ungskogsstammar och röjningsbehov med bra resultat, men de har inte övertygat företagen att lämna de traditionella fältundersökningarna som uppfyller kraven på noggrannhet och hittills har varit konkurrenskraftiga i kostnadskalkylen. Högupplösta fotogrammetriska punktmoln och ALS-data kan ge noggranna uppskatt-ningar av ungskogens biofysiska egenskaper men kräver också intensiv databehand-ling. Ett sätt att minska kostnaderna kan vara att använda mindre exakta 2D-bilder genom att endast skala bilder till den approximerade flyghöjden. Denna studie har utvärderat grunderna för att använda enkla drönarbilder till in-ventering av boreala ungskogar och resultaten jämfördes med Sveaskogs nuvarande fältbaserade inventeringspraxis av unga skogsbestånd. Syftet var att utvärdera precis-ionen i tre huvudfrågor: ’Kamerans positionering och bildens skala’, Trädslags- och stamantals klassificering’ samt ’Bedömt röjningsbehov på beståndsnivå’. Studieom-rådet omfattade 57 bestånd och 290 provytor i närheten av Vindeln, Västerbottens län. Materialet i studien och det praktiska tillvägagångssättet bestämdes huvudsakli-gen utifrån tillgänglig utrustning och fältinstruktioner erhållna från Sveaskog. Resultaten visade att manuell tolkning av de förvärvade bilderna kunde skilja mel-lan skogsbestånd som behöver röjas och inte med 82% noggrannhet, även om tolkat antal stammar per provyta omfattade stora fel (t.ex. RMSE 6568 för totalt antal stam-mar ha-1). Möjliga fördelar jämfört med traditionella inventeringar är; framställning av uppdaterade kartor av hög kvalitet för röjningsarbetarna, förbättrad planering och avgränsning av hänsyn och framför allt minskad tid i fält eftersom över 1 miljon hek-tar beräknas vara i behov av röjning i Sverige årligen. Denna studie indikerar att enkla drönarbilder utan korrekt fotogrammetri kan vara ett alternativ till dagens manuella fältinventeringar. Kvalitén av bilderna kan vara låg men möjligen tillräcklig. 2020-02-17 Second cycle, A2E NonPeerReviewed application/pdf sv https://stud.epsilon.slu.se/15343/15/kingstad_v_%20200212.pdf Kingstad, Victor, 2020. Evaluated density estimates of young forest stands using high resolution 2D imagery from UAV. Second cycle, A2E. Umeå: (S) > Dept. of Forest Resource Management <https://stud.epsilon.slu.se/view/divisions/OID-260.html> urn:nbn:se:slu:epsilon-s-15343 eng |
| spellingShingle | Forestry - General aspects Forestry production Kingstad, Victor Evaluated density estimates of young forest stands using high resolution 2D imagery from UAV |
| title | Evaluated density estimates of young forest stands using high resolution 2D imagery from UAV |
| title_full | Evaluated density estimates of young forest stands using high resolution 2D imagery from UAV |
| title_fullStr | Evaluated density estimates of young forest stands using high resolution 2D imagery from UAV |
| title_full_unstemmed | Evaluated density estimates of young forest stands using high resolution 2D imagery from UAV |
| title_short | Evaluated density estimates of young forest stands using high resolution 2D imagery from UAV |
| title_sort | evaluated density estimates of young forest stands using high resolution 2d imagery from uav |
| topic | Forestry - General aspects Forestry production |
| url | https://stud.epsilon.slu.se/15343/ https://stud.epsilon.slu.se/15343/ |