Mare Humboldtianum is a Nectarian-aged basin on the northeast limb of the Moon, whose inner ring was flooded by basalt billions of years ago. The topography of Mare Humbold is rather complex, which makes it difficult to define the basaltic units only by the spectral data. To improve understanding of the geological units in Mare Humbold-28 tianum and provide new interpretations of the formation process, we use the brightness temperature (TB) data obtained by the Chang’E-2 (CE-2) mission, which is proved to have a good penetration feature. The TB difference (dTB) maps at multichannels are created to analyze the TB variations. The Backpropagation Neural Network (BPNN) model is first used to provide a statistical interpretation of the relation between dTB distributions and surface parameters. Our results show that, most of the dTB performances are strongly affected by the topography. Several abnormally dTB performances are observed on dTB maps. This observation indicates that there exists two different mare units in the eastern and western parts in the basin floor. These TB variations are interpreted that most of the dTB anomalies are strongly related to the special tectonic features. The statistical results based on BPNN method indicate that multichannel TB variations reflect different topographic features. Most of the topographic effects related to dTB variations are found in the regions controlled by the small craters and topographic bulges in the basin floor.