Spatial dependence between training and test sets: another pitfall of classification accuracy assessment in remote sensing.
Nicolas KarasiakJean-Francois DejouxClaude MonteilDavid SheerenPublished in: Mach. Learn. (2022)
Keyphrases
- remote sensing
- remote sensing data
- classification accuracy
- training and test sets
- change detection
- spatial analysis
- multispectral
- remote sensing images
- remotely sensed data
- geographical information systems
- remote sensing imagery
- high resolution
- training set
- image analysis
- image processing
- satellite images
- high spatial resolution
- image fusion
- spatial data
- satellite imagery
- satellite data
- spatial information
- feature set
- automatic image registration
- test set
- hyperspectral
- multi spectral images
- training data
- support vector machine
- earth science
- generalization error
- hyperspectral images
- digital image analysis
- spatial relationships
- naive bayes
- hyperspectral remote sensing
- earth observation
- hyperspectral imagery
- naive bayes classifier
- spatial databases
- sensor networks
- feature selection
- land cover
- spatial resolution
- pattern recognition
- support vector
- decision trees
- machine learning
- data sets