Building recognition with partial 3D-information collected by airborne sensors at limited angles

A remote sensing (RS) building recognition frame using partial 3D-information is proposed in this paper. Since there are always some constraints such as launch capability of aircrafts, airspace rights, capability of remote sensors and etc, RS information (2D images or 3D data) of buildings is normally collected at some limited angles by airborne sensors, and owing to self-occlusion, only partial information could be obtained. It is extremely difficult to distinguish variety buildings correctly merely by partial information. This paper proposes a novel method semantic multi-view-split (SMVS) model for analyzing the style of partial 3D-information. Then 3D-Zernike Descriptors are employed as the feature of different 3D partial data owing to their excellent rotation invariance and robustness. Experiments show that a high recognition rate could be obtained by our frame.