Research on Ransac Filtering Optimization Method of Steel Structure Buildings Based on Morphological Features

Abstract

The laser point cloud data of complex steel structure buildings and structures includes a large number of noise points, and because the shielding is serious, it is easy to lead to the extremely uneven point density, leading to serious impacts on accurate noise elimination. Traditional filtering methods commonly use a single mathematical model, which is difficult to achieve effective filtering. In this paper, a filtering method combining RANSAC algorithm with point cloud principal component analysis and connectivity analysis is proposed. Firstly, according to the cylinder structural characteristics of the steel structure, RANSAC algorithm is used to detect the cylinder point cloud, eliminating some scattered noise points. Then the principal component analysis is carried out to obtain the noise points appeared as a small number of clusters. Finally, according to the connectivity characteristics of the structure, the point cloud connectivity is used to distinguish the noise point cluster from the steel structure point cloud, so as to obtain a more accurate and complete steel structure point cloud. In this paper, the laser point cloud of a sports venue is taken as the experimental data, processed by the proposed filtering method, and then compared with the traditional filtering method. The experimental results show that the filtering method can eliminate the noise more accurate and obtain a more complete steel structure point cloud.