Indoor 3D wireframe construction from incomplete point clouds based on Gestalt rules

Abstract

The complexity of indoor point cloud and information loss during point cloud acquisition are the main challenges for accurate and complete indoor buildings 3D wireframe construction. This paper presents a 3D wireframe construction method for indoor structure point clouds based on Gestalt rules. First, boundary points of indoor structures are used for structural plane segmentation, including ceilings, floors, walls, beams, etc. Then, point clouds of each structural plane are projected onto its best-fit plane to obtain a 2D projection, followed by boundary points extraction and line fitting to get the 2D line segments. Finally, an accurate and complete 3D wireframe is constructed by optimizing anomalous line segments based on the Gestalt rules. In our method, Gestalt rules are used to deal with problems in line segments, including cluttered distribution, duplicate extraction, non-regularization (non-parallel and non-perpendicular), topological errors (non-closed and discontinuous), and missing information. Experiments demonstrate that our proposed method significantly improves the accuracy and completeness of the results.