Item

Abstract

Inadequate and inhomogeneous GNSS (Global Navigation Satellite System) observation severely restricts the advancement of the voxel-based computerized ionospheric tomography (CIT) technique. Various inversion algorithms, exemplified by the algebraic reconstruction technique (ART), have been proposed for estimating ionospheric electron density (IED). However, using ART techniques, the IEDs of voxels that are not crossed by GNSS rays cannot be accurately estimated. In this study, we proposed a compressed sensing technique (CST) for ionospheric tomography using the K-SVD algorithm for dictionary learning. Given a set of IED values that are derived from the NeQuick-2 model, the K-SVD algorithm uses the orthogonal matching pursuit for sparse coding and the singular value decomposition (SVD) approach for dictionary updating. When compared to the multiplicative algebraic reconstruction method (MART) algorithm, both simulation and real experiments proved the viability and superiority of the CST technique. When the NeQuick-2 model was employed for dictionary learning, it was discovered that the tomographic performances of the CST method were better than those of the MART algorithm. Besides, the tomographic results were used for dictionary learning in CST-based tomography. It showed that the CST algorithm can further improve tomographic performances when compared to the MART algorithm. Additionally, there was a better agreement between the IED profiles derived from radio occultation data and the electron density profiles produced by the CST method.