中外合作获得高精度月表化学成分分布图

论文原文说得那么清楚了：“In this work, the MI spectral features and the measured oxide abundances from sample-return sites both by adding Chang’e-5 samples and by only with Apollo and Luna data (w/o Chang’e-5) have been used to build two training sets to derive the inversion model.”
"MI spectral features"是日本的“月女神”探测器（SELENE (KAGUYA) multiband imager (MI)）的八个紫外-可见-近红外波段光谱数据(UV–vis and NIR spectroscopy with eight wavelengths)，用嫦娥五号、阿波罗和月球着陆采样的样品实测数据，“采用深度学习方法，精确反演了月球表面主要元素的含量”，也就是把化学成分实测值与光谱数据做了对应，建立了光谱特征与实际成分之间关系的数学模型（“校准”只是简单通俗的说法），从而利用月球全球光谱数据推算出月球全球的元素（对应的氧化物）分布情况。