Convolutional neural networks and stochastic modelling in hyperspectral data analysis

Abstract

Hyperspectral imaging is relatively new and rapidly growing field of research. The datasets produced by hyperspectral imaging are large, and handling such data requires large computational resources. Therefore, there is a need for developing machine learning methods that can cope with the data, and methods to reduce the necessary amount of data gathering missions. For the latter, problem the author and his co-authors have developed stochastic modelling and generative adversarial neural networks for data augmentation. In machine learning, they have experimented with using convolutional neural network in conjunction with said stochastic model in order to retrieve useful information from hyperspectral data. Additionally, the author lists useful Python packages for hyperspectral data analysis. Keywords: Hyperspectral imaging, Convolutional neural network, Stochastic modelling, Biophysical parameter retrieval, Data augmentation