Journal
ADVANCES IN SPACE RESEARCH
Volume 67, Issue 7, Pages 2031-2043Publisher
ELSEVIER SCI LTD
DOI: 10.1016/j.asr.2021.01.004
Keywords
Space object; HSI; Material identification; Tensor decomposition
Categories
Funding
- strategic priority research program of the Chinese Academy Sciences [XDA17040203, XDA19010103]
- National natural science foundation of China [21827808]
Ask authors/readers for more resources
A space object material identification method based on Tucker decomposition is proposed and demonstrated experimentally in this study, showing a superior performance compared to other decomposition methods.
Space object material identification method based on Tucker decomposition is proposed and demonstrated experimentally. Space target generally has a low spatial resolution because of the limitation in detection distance. Hyperspectral imaging (HSI) technology can capture the image from visible light to shortwave infrared continuously while providing the spatial and spectral information of the target, thereby introducing a new space target detection and identification approach. However, the data obtained by the HSI system often contain mixed pixels, thereby causing difficulties in target material identification. In this work, a material identification method of hyperspectral data based on Tucker decomposition is proposed by combining mixed spectral theory with tensor decomposition. The feasibility of the method is verified by using satellite model hyperspectral data with different spatial resolutions compared with the endmember obtained from the non-negative matrix decomposition (NMF), independent component analysis (ICA), tensor singular value decomposition (t-SVD). The average CORR of NMF, ICA, t-SVD and the proposed method is 0.2694, 0.5818, 0.6397 and 0.937, correspondingly. Therefore, the proposed method has demonstrated a more remarkable performance in terms of material identification, the analysis results of the material abundance distribution that used the proposed method. (C) 2021 COSPAR. Published by Elsevier B.V. All rights reserved.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available