2020/07/12 by John Janiczek, Janiczek, John, Parth Thaker +9 · 1 citation
Engineering · Environmental Science · #Advanced Image Fusion Techniques #Atmospheric and Oceanic Physics (physics.ao-ph) #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #FOS: Electrical engineering #FOS: Physical sciences #Image and Video Processing (eess.IV) #Remote Sensing in Agriculture #Remote-Sensing Image Classification #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.2007.05996
openalex publication_date 2020/07/12 · openalex created_date 2022/07/26 · openalex updated_date 2026/07/28
Hyperspectral unmixing is an important remote sensing task with applications\nincluding material identification and analysis. Characteristic spectral\nfeatures make many pure materials identifiable from their visible-to-infrared\nspectra, but quantifying their presence within a mixture is a challenging task\ndue to nonlinearities and factors of variation. In this paper, spectral\nvariation is considered from a physics-based approach and incorporated into an\nend-to-end spectral unmixing algorithm via differentiable programming. The\ndispersion model is introduced to simulate realistic spectral variation, and an\nefficient method to fit the parameters is presented. Then, this dispersion\nmodel is utilized as a generative model within an analysis-by-synthesis\nspectral unmixing algorithm. Further, a technique for inverse rendering using a\nconvolutional neural network to predict parameters of the generative model is\nintroduced to enhance performance and speed when training data is available.\nResults achieve state-of-the-art on both infrared and visible-to-near-infrared\n(VNIR) datasets, and show promise for the synergy between physics-based models\nand deep learning in hyperspectral unmixing in the future.\n