Blended learning for hyperspectral data

Ilya Kavalerov; Wojciech Czaja

doi:10.1117/12.2519977

14 May 2019 Blended learning for hyperspectral data

Ilya Kavalerov, Wojciech Czaja

Proceedings Volume 10986, Algorithms, Technologies, and Applications for Multispectral and Hyperspectral Imagery XXV; 109861B (2019) https://doi.org/10.1117/12.2519977
Event: SPIE Defense + Commercial Sensing, 2019, Baltimore, MD, United States

Abstract

We explore the spectral spatial representation capabilities of convolutional neural networks for the purpose of classification of hyperspectral images. We examine several types of neural networks, including a novel technique that blends the Fourier scattering transform with a convolutional neural network. This method is naturally suited for the representation of hyperspectral data because it decomposes signals into multi-frequency bands, removing small perturbations such as noise, while also having the capability of neural networks to learn a hierarchical representation. We test our proposed method on the standard Pavia University hyperspectral dataset and demonstrate a new training set sampling strategy that reveals the inherent spatial bias present in some purely neural network methods. The results indicate that our form of blended learning is more effective at representing spectral data and less prone to overfitting the artificial spatial bias in hyperspectral data.

Citation Download Citation

Ilya Kavalerov and Wojciech Czaja "Blended learning for hyperspectral data", Proc. SPIE 10986, Algorithms, Technologies, and Applications for Multispectral and Hyperspectral Imagery XXV, 109861B (14 May 2019); https://doi.org/10.1117/12.2519977

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available