14 January 2021 in IEEE Transactions on Geoscience and Remote Sensing
Band selection (BS) is a commonly used dimension reduction technique for hyperspectral images. In this article, we propose a similarity-based ranking (SR) strategy inspired by a density-based clustering algorithm. The representativeness of a band is evaluated according to its ability to become a cluster center. We introduce structural similarity (SSIM) to measure the relationships between the bands. Thus, our proposed ranking-based BS method is called SR-SSIM. We picked state-of-the-art BS methods as competitors and carried out classification experiments on different data sets. The results illustrated that SR-SSIM outperformed the other methods. It is demonstrated, in this article, that the SSIM is more suitable for hyperspectral BS than the Euclidean distance since the SSIM can mine the spatial information contained in the band images. Furthermore, we discuss the application of BS methods on deep learning classifier. We found that proper preprocessing by the BS method can effectively eliminate redundant information and avoid overfitting.
28 April 2022 in Optics Express
Most of the existing deep learning methods for hyperspectral image (HSI) classification use pixel-wise or patch-wise classification. In this paper, we propose an image-wise classification method, where the network input is the original hyperspectral cube rather than the spectral curve of each pixel (i.e., pixel-wise) or neighbor region of each pixel (i.e., patch-wise). Specifically, we propose a minimalistic fully convolution network (MFCN) and a semi-supervised loss function, which can perform pixel-level classification for HSI with few labeled samples. The comparison experiments demonstrated the progress of our methods, using three new benchmark HSI datasets (WHU-Hi-LongKou, WHU-Hi-HanChuan and WHU-Hi-HongHu) with wavelength range from 400 to 1000nm. In the comparison experiments, we randomly selected 25 labeled pixels from each class for training, equivalent to only 0.11%, 0.16%, and 0.14% of all labeled pixels for the three datasets, respectively. In addition, through ablation studies and theoretical analysis, we verified and analyzed the effectiveness and superiority of our design choices.
16 February 2021 in Journal of Geophysical Research: Space Physics
The low‐latitude longitudinally periodic density structure with three or four peaks at a fixed local time is well known. Both harmonic analysis of observations and numerical simulations have been extensively performed to ascribe the periodic structures to different atmospheric tides. But to date, no research has been done to explain how the two structures transform between seasons. Besides, other different wave structures have also been reported. To clarify these uncertainties, the COSMIC electron density data during 2006–2019 are used to further examine the spatiotemporal variations of the longitudinal structure. The results show that a wavenumber‐4 structure is clearly distinguishable during equinoxes and the four peaks are basically the same size, but in two solstices, a wavenumber‐3 structure is finally observed and the three peaks vary greatly in size. The wavenumber‐4 is well established before noon while the stable wavenumber‐3 form cannot be reached until after noon. Due to the blurred boundary caused by plasma transportation, the density distribution can be characterized by other wave structures before reaching its stable form during solstices. Subdividing the data, we find that from vernal (autumn) equinox to June (December) solstice, the peak in American sector gradually merges with its west (east) side peak into a super‐large one, its east (west) side peak weakens into a small one, and the peak in Southeast Asian sector gradually shrinks (expands). We speculate that the transition from wavenumber‐4 to wavenumber‐3 is due to the modulation of transequatorial winds and magnetic declination.