Abstract
In this study, a new approach is described in hyperspectral remote sensing image classification problem. In this approach, each band of hyperspectral image is decomposed to details by using à trous wavelet decomposition method. The obtained details are injected to the related hyperspectral band additively and proportionally. Thus, the hyperspectral image de- tails are improved. The resulting new image after dimensionality reduction is trained and classified with conventional classifiers and ensemble methods, and their performance is discussed. It has been proven that the proposed method improves the metric results in terms of the overall accuracy and Kappa.
Keywords
Hyperspectral imaging
Artificial intelligence
Pattern recognition (psychology)
Dimensionality reduction
Wavelet
Computer science
Image (mathematics)
Curse of dimensionality
Metric (unit)
Decomposition
Wavelet transform
Contextual image classification
Computer vision
Mathematics
Engineering
Chemistry
Subject Areas
Remote-Sensing Image Classification
·Media Technology ·Physical Sciences
Remote Sensing and Land Use
·Atmospheric Science ·Physical Sciences
Advanced Image Fusion Techniques
·Media Technology ·Physical Sciences
Loading…