Transformation of Multispectral Data to Quasi-Hyperspectral Data Using Convolutional Neural Network Regression

Hyperspectral (HS) data are proven to be more resourceful compared multispectral (MS) for object detection, classification, and several other applications. However, absence of any space-borne HS sensor since 2017, which can provide open-source with global coverage, high cost limited obtainability airborne sensors-based images limit the use data. Transformation readily available MS into quasi-HS a feasible solution this issue. In article, we propose convolutional neural network regression (CNNR), deep learning-based algorithm, (i.e., Landsat 7/8) quasi-Hyperion) transformation. The proposed CNNR model is existing pseudo-HS image transformation algorithm (PHITA), simple linear [i.e., stepwise (SLR)], nonlinear modeling approach support vector (SVR)] by evaluating quality quasi-Hyperion Contrary these models, has added advantage utilizing spectral-spatial features through regression-based modeling. Different statistical metrics calculated compare each band's reflectance values as well spectral curve pixel that original Hyperion developed models generated also evaluated application crop classification. Analyzing results all experiments, it evident efficient PHITA, SLR, SVR in creating transformed classification application. model-based used viable alternative different applications images.