Logistic Regression-Based Spectral Band Selection for Tree Species Classification: Effects of Spatial Scale and Balance in Training Samples.
Paras PantVille HeikkinenIlkka KorpelaMarkku Hauta-KasariTimo TokolaPublished in: IEEE Geosci. Remote. Sens. Lett. (2014)
Keyphrases
- training samples
- band selection
- pixel classification
- training set
- feature space
- hyperspectral images
- hyperspectral imagery
- supervised learning
- hyperspectral data
- hyperspectral
- learning algorithm
- hyperplane
- test sample
- training data
- hyperspectral remote sensing
- high dimensional
- spectral bands
- multispectral
- spatial information
- support vectors
- image classification
- decision trees
- gaussian process
- remote sensing
- principal component analysis
- classification accuracy
- active learning
- feature extraction