Deeply learned attribute profiles for hyperspectral pixel classification

Official URL: http://risc01.sabanciuniv.edu/record=b1640326 (Table of Contents)

Hyperspectral Imaging has a large potential for knowledge representation about the real world. Providing a pixel classi cation algorithm to generate maps with labels has become important in numerous elds since its inception, found use from military surveillance and natural resource observation to crop turnout estimation. In this thesis, within the branch of mathematical morphology, Attribute Pro les (AP) and their extension into the Hyperspectral domain have been used to extract descriptive vectors from each pixel on two hyperspectral datasets. These newly generated feature vectors are then supplied to Convolutional Neural Networks (CNNs), from o -the-shelf AlexNet and GoogLeNet to our proposed networks that would take into account local connectivity of regions, to extract further, higher level abstract features. Bearing in mind that the last layers of CNNs are supplied with softmax classi ers, and using Random Forest (RF) classi ers as a control group for both raw and deeply learned features, experiments are made. The results showed that not only there are signi cant improvements in numerical results on the Pavia University dataset, but also the classi cation maps become more robust and more intuitive as di erent, insightful and compatible attribute pro les are used along with spectral signatures with a CNN that is designed for this purpose.