Abstract:

The line features extracted from point clouds are very useful in the processing of point clouds, including symmetry detection, surface reconstruction, the registration from image to point clouds, etc. However, the ability of existing line feature extraction approaches to deal with noise, outliers, and missing parts in the data is limited. On the other hand, RANSAC (RANdom SAmpling Consensus) based methods are widely used in the fields of image processing and 3D model processing because of the robustness. Thus, a RANSAC based line features detection algorithm is proposed in this paper, in which RANSAC is first used to detect all the possible planes in the point clouds, then to detect the line features from the boundary points of the parameterization field of the planes with global constraints. This method is designed especially for point clouds obtained from architectures or mechanical parts, in which planar features are dominant. Result of experiments validates the robustness of the proposed algorithm in handling with various defections of point clouds, e.g. noise, outliers, and data missing.

Key words: point clouds, line feature, RANSAC, robust