Geochemical classification of lunar rocks is mainly investigated through two methods: laboratory analyses of lunar rock or meteorite samples to determine material composition, and the use of multi-source planetary remote sensing spectral data to infer the composition of lunar surface materials. Standards for classifying lunar rocks based on remote sensing spectra and geochemical analyses of returned lunar samples vary, each presenting unique advantages and limitations. Employing any single approach alone is insufficient for devising a precise geochemical classification scheme for lunar rock types. This study conducts a geochemical classification analysis of mare basalts in the Mare Crisium region by using Clementine UV/VIS multispectral remote sensing data, based on the differences in inferred TiO2 content. Consequently, the mare rocks in this region are categorized into five main types. Through further investigations into the lithological, chronological, and structural characteristics of both the shallow and deep layers of Mare Crisium, this study highlights the importance of using a multi-source data integration approach that combines lunar remote sensing with lunar rock sample data to jointly constrain lunar rock classification schemes. This integration provides new lunar surface truth values for quantitative inversion of lunar remote sensing data, enabling a more scientific approach to geochemical classification studies of lunar rocks using lunar remote sensing data.