Geoscience features extraction plays an important role in the quantitative analysis of mathematical geology, which provides an objective basis for identifying the types of surface units. In this paper, a new type of Geoscience features is defined to characterize the roughness of surfaces. In addition, the mathematical characteristic of the new feature is discussed. The new method relies on the theory of stochastic processes. Both the spectrum moments and its statistical invariants are considered as the indicators for characterizing the surface physiography. The experiments are performed on both the satellite gravity data in China and Digital Elevation Model (DEM) data. And the process of geomorphology recognition and steps for the calculation process are demonstrated. The results show that the new feature can reflect the relief characteristics of landforms, and the method can also be applied to create Gravity tectonic units and segmentationbased maps of physiography.