A Machine Learning Approach for Material Detection in Hyperspectral Images

摘要

In this paper we propose a machine learning approachfor the detection of gaseous traces in thermal infra redhyperspectral images. It exploits both spectral and spatialinformation by extracting subcubes and by using extremelyrandomized trees with multiple outputs as a classifier.Promising results are shown on a dataset of morethan 60 hypercubes.