Abstract: Aiming at the problems of traditional feature extraction algorithms, such as cross disorder, discontinuity and lack of topological relationship, this paper proposes a point cloud feature extraction algorithm based on Morse theory. Firstly,the Morse function index of the vertices on the 3D surface model is defined. Secondly, the adjacent point comparison method is used to extract the feature points automatically. Finally, aiming at the problem that the duality of Morse-Smale complex has no practical significance in building topology features, a single complex topology model extraction and simplification algorithm is proposed. The experimental results show that the algorithm can obtain clear, continuous and complete building feature lines and realize the complete segmentation of the building model surface. On the premise of ensuring the topological consistency of building feature lines, the simplified algorithm can obtain different levels of building topology features, which provides a guarantee for the reconstruction and visualization of building models.

Key words: building point clouds, feature extraction, Morse theory, single complex, feature simplification