期刊
ICARUS
卷 321, 期 -, 页码 200-215出版社
ACADEMIC PRESS INC ELSEVIER SCIENCE
DOI: 10.1016/j.icarus.2018.10.031
关键词
Moon; Oxide; Neural network
资金
- National Natural Science Foundation of China [41372341]
- National Key Project [2016YFC0702000]
- National Students' Platform for Innovation and Entrepreneurship Training Program [201710491048]
The major oxides (SiO2, Al2O3, CaO, FeO, MgO, and TiO2) and Mg# are critical for revealing the petrological characteristics of the Moon and for testing models of lunar formation and geologic evolution. There are few high-spatial-resolution (< 250 m/pixel) abundance maps for all the six major oxides and Mg# across the Moon. Furthermore, previous studies primarily employed the traditional regression methods to derive oxide contents from optical images, which may influence the inversion accuracies of the lunar chemical compositions. This paper reports the abundance maps of all the six major oxides and Mg# with a relatively high spatial resolution of similar to 200 m/pixel and compared them with the ones in the previous works. Neural networks algorithms along with the data from the Interference Imaging Spectrometer (IIM) onboard Chang'E-1 were employed in this paper to derive the abundances of the six oxides. Compared with the traditional linear regression models, the neural networks method suggested in this work is hopeful to better depict the complex nonlinear relations between the spectra and the chemical components, so it may improve the inversion performance of the lunar chemistry.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据