Deep learning and machine learning for hyperspectral imaging

Hyperspectral imaging captures in an image all the information of the electromagnetic spectrum generating valuable data for a wide variety of applications, my research focuses on infrastructural monitoring. Hyperspectral sensors acquire high dimensional data which Machine Learning (ML) and Deep Learning (DL) architectures can manage but require high computational sources. The aim is to exploit the advantages of using ML and DL for hyperspectral image classification, while reducing the computing time by applying dimensionality reduction to input data. An approach using the Wavelet transform is described in the following. Wavelet features are then given as input to a cubic Support Vector Machine (SVM) and a 3D Convolutional Neural Network (CNN).