March 1, 1996International Journal of Remote Sensing
Classification trees: an alternative to traditional land cover classifiers
Authors
Discussion
Loading...
Member takes
Implication
Comparative analysis reveals classification trees achieve 82 percent accuracy in global land cover classification, suggesting strong utility for non-parametric and high-dimensional datasets.
Key Points
- To evaluate classification trees as a flexible, hierarchical alternative to traditional maximum likelihood classifiers for mapping global land cover types.
- Compared a hierarchical decision classification tree against a maximum likelihood classifier.
- Trained and tested classification performance on a 1° by 1° global validation dataset.
- The classification tree achieved 82% accuracy on the validation dataset, matching the 82% accuracy observed with the maximum likelihood classifier.
- The classification tree successfully reduced dataset dimensionality and identified the most informative metrics for distinguishing land cover classes.
Cite This Study
Hansen et al. (1996) studied this question.