University of Twente Student Theses

Login
This website will be unavailable due to maintenance December 1st between 8:00 and 12:00 CET.

Tree species classification using seasonal UAV-RGB data, summer UAV multi-spectral data, and its combination in a mixed temperate forest: a case study of Haagse Bos, Netherlands.

Addo, Efia Boakye (2021) Tree species classification using seasonal UAV-RGB data, summer UAV multi-spectral data, and its combination in a mixed temperate forest: a case study of Haagse Bos, Netherlands.

Link to full-text:https://ezproxy2.utwente.nl/login?url=https://library.itc.utwente.nl/login/2021/msc/nrm/addo.pdf
(only accessible for UT students and staff)
Abstract:Tree species classification is of importance to various groups of people who manage and protect the forest. While some forest managers need this information to undertake the appropriate silvicultural activities, others need this information to identify tree species that can be harvested for timber. Tree species information can be used to identify and conserve endangered species, indicates biodiversity richness, and is also used to identify a species-specific allometric equation to estimate forest biomass and carbon. Although tree species information can be obtained at any time, the tree’s phenological changes make it possible to identify certain tree species better in a specific season than the others. In view of this, this research aimed to compare the results of classification of tree species using seasonal UAV-RGB images, summer UAV-MSS images, and its combination in a temperate forest. The object-based image analysis and support vector machine classification algorithm was deployed to classify eight tree species and a combination of tree species named others. The tree crowns were first identified using multi-resolution segmentation (MRS) algorithm in object-based image analysis. The accuracy of the tree crowns was assessed by estimating the under segmentation, over-segmentation, and overall segmentation error. The generated tree crowns, combined with tree species data collected from the field, were used to classify the seasonal UAV-RGB datasets, summer UAV-MSS dataset, and the combination of summer UAV-RGB and MSS datasets using a support vector machine classification algorithm. The accuracy of the tree species classification was then assessed by generating the confusion matrix. The summer MSS dataset produced a segmentation accuracy of 82%, which was the highest among the other seasonal dataset. The summer RGB, fall RGB, summer MSS and RGB, spring, and winter RGB yielded segmentation accuracy of 76%, 73%, 71%, 57%, and 31%, respectively. The result of the tree species classification showed that the summer UAV-RGB dataset resulted in the highest classification accuracy of 0.77 compared to the winter, fall, and spring UAV-RGB datasets, resulting in the overall accuracy of 0.49, 0.68, and 0.64, respectively. The summer UAV-MSS dataset yielded an overall accuracy of 0.84, while the combination of summer UAV-RGB and MSS yielded the best overall accuracy of 0.88. This research suggests that combining UAV-RGB and MSS datasets of the same season can improve tree species classification, better estimating species-level above-ground biomass and sustainable management of natural resources.
Item Type:Essay (Master)
Faculty:ITC: Faculty of Geo-information Science and Earth Observation
Programme:Geoinformation Science and Earth Observation MSc (75014)
Link to this item:https://purl.utwente.nl/essays/88981
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page