Mapping grazed vegetation communities on Macquarie Island using a binary ensemble classifier.
This study implemented and applied a binary ensemble classifier for identification of grazed vegetation communities on Macquarie Island from very high resolution Quickbird imagery. Rabbit grazing has severely affected Macquarie’s unique sub-Antarctic vegetation communities. The aim of this study was...
| Main Author: | |
|---|---|
| Format: | Conference Object |
| Language: | English |
| Published: |
2008
|
| Subjects: | |
| Online Access: | https://eprints.utas.edu.au/6980/ https://eprints.utas.edu.au/6980/1/14arspc_lucieer_224.pdf http://www.14arspc.com/ |
| Summary: | This study implemented and applied a binary ensemble classifier for identification of grazed vegetation communities on Macquarie Island from very high resolution Quickbird imagery. Rabbit grazing has severely affected Macquarie’s unique sub-Antarctic vegetation communities. The aim of this study was to identify the grazed areas from Quickbird imagery to map their spatial extent. Seven different soft classification algorithms were applied to classify the image into grazed vs. ‘other’ classes. The maximum likelihood classifier, supervised fuzzy c-means classifier (Euclidean distance, Mahalanobis distance, and k-nearest neighbour), and three support vector machine classifiers (SVM) were applied. An ensemble classifier based on the consensus rule was used to combine the seven classification results. A very high classification accuracy of 97% was achieved with the ensemble classifier, identifying grazed areas and providing an estimate of classification uncertainty. |
|---|