Ancient architecture point cloud data segmentation based on modified fuzzy C-means clustering algorithm

Abstract

Segmentation of Point cloud data is a key but difficult problem for architecture 3D reconstruction. Because compared to reverse engineering, there are more noise in ancient architecture point cloud data of edge because of mirror reflection and the traditional methods are hard that is not fuzzy in the preceding part of this paper, these methods can't embody the case of the points of borderline belonging two regions and it is difficult to satisfy demands of segmentation of ancient architecture point cloud data. Ancient architecture is mostly composed of columniation, plinth, arch, girder and tile on specifically order. Each of the component's surfaces is regular and smooth and belongingness of borderline points is very blurry. According to the character the author proposed a modified Fuzzy C-means clustering (MFCM) algorithm, which is used to add geometrical information during clustering. In addition this method improves belongingness constraints to avoid influence of noise on the result of segmentation. The algorithm is used in the project "Digital surveying of ancient architecture--- Forbidden City". Experiments show that the method is a good anti-noise, accuracy and adaptability and greater degree of human intervention is reduced. After segmentation internal point and point edge can be districted according membership of every point, so as to facilitate the follow-up to the surface feature extraction and model identification, and effective support for the three-dimensional model of the reconstruction of ancient buildings is provided.