DETECTION OF HYPERSPECTRAL IMAGES

Abstract

Hyperspectral imaging is a powerful technique that captures data from a wide range of
electromagnetic wavelengths, enabling the detailed analysis of materials and objects. However, processing
hyperspectral images posses significant challenges due to their high dimensionality and complex spectral
information. In recent years, terse representation and fully connected neural networks have emerged as
effective tools for hyperspectral image analysis. This paper proposes Detection of Hyperspectral images using
terse representation and Fully connected neural networks. This system uses the advantages of terse
representation and fully connected neural networks. At first the hyperspectral images are converted into a
compact terse representation reducing its dimensionality and preserving the spectral information and then
this is feed to fully connected neural networks. These networks are trained to learn the underlying patterns
present in the data. The result obtained by combination of terse representation and fully connected neural
networks gives better accuracy than traditional methods