Abstract:As-built Building Information Models (BIMs) are becoming increasingly popular in the Architectural, Engineering, Construction, Owner and Operator (AECOO) industry. These models reflect the state of the building up to as-built conditions. The production of these models for existing buildings with no prior BIM includes the segmentation and classification of point cloud data and the reconstruction of the BIM objects. The automation of this process is a must since the manual Scan-to-BIM procedure is both time-consuming and error prone. However, the automated reconstruction from point cloud data is still ongoing research with both 2D and 3D approaches being proposed. There currently is a gap in the literature concerning the quality assessment of the created entities. In this research, we present the empirical comparison of both strategies with respect to existing specifications. A 3D and a 2D reconstruction method are implemented and tested on a real life test case. The experiments focus on the reconstruction of the wall geometry from unstructured point clouds as it forms the basis of the model. Both presented approaches are unsupervised methods that segment, classify and create generic wall elements. The first method operates on the 3D point cloud itself and consists of a general approach for the segmentation and classification and a class-specific reconstruction algorithm for the wall geometry. The point cloud is first segmented into planar clusters, after which a Random Forests classifier is used with geometric and contextual features for the semantic labelling. The final wall geometry is created based on the 3D point clusters representing the walls. The second method is an efficient Manhattan-world scene reconstruction algorithm that simultaneously segments and classifies the point cloud based on point feature histograms. The wall reconstruction is considered an instance of image segmentation by representing the data as 2D raster images. Both methods have promising results towards the reconstruction of wall geometry of multi-story buildings. The experiments report that over 80% of the walls were correctly segmented by both methods. Furthermore, the reconstructed geometry is conform Level-of-Accuracy 20 for 88% of the data by the first method and for 55% by the second method despite the Manhattan-world scene assumption. The empirical comparison showcases the fundamental differences in both strategies and will support the further development of these methods.

Volumetric wall detection in unorganized indoor point clouds using continuous segments in 2D grids

Volumetric Representation and Sphere Packing of Indoor Space for Three-Dimensional Room Segmentation

Reconstruction of Indoor Navigation Elements for Point Cloud of Buildings with Occlusions and Openings by Wall Segment Restoration from Indoor Context Labeling

Detection of doors in a voxel model, derived from a point cloud and its scanner trajectory, to improve the segmentation of the walkable space

A Specialized Pipeline for Efficient and Reliable 3D Semantic Model Reconstruction of Buildings from Indoor Point Clouds

Comparison of 2D and 3D wall reconstruction algorithms from point cloud data for as-built BIM

A New Framework for Generating Indoor 3D Digital Models from Point Clouds

Semantic-aware Room-Level Indoor Modeling from Point Clouds

Surface Extraction and Boundary Detection Based on DBSCAN Clustering in 3D Point Clouds

Scene Segmentation and Understanding for Context-Free Point Clouds

Automated Semantic Segmentation of Indoor Point Clouds from Close-Range Images with Three-Dimensional Deep Learning

Boundary-Aware Supervoxel Segmentation for Indoor 3D Point Clouds

Geometry and Topology Reconstruction of BIM Wall Objects from Photogrammetric Meshes and Laser Point Clouds

Geometrically consistent plane extraction for dense indoor 3D maps segmentation

The Polygonal 3D Layout Reconstruction of an Indoor Environment via Voxel-Based Room Segmentation and Space Partition

3D Semantic Segmentation Using Deep Learning for Large-Scale Indoor Point Cloud

AUTOMATIC EXTRACTION OF FACADES AND WINDOWS FROM MLS POINT CLOUDS USING VOXELSPACE AND VISIBILITY ANALYSIS

3D Surface Segmentation from Point Clouds Via Quadric Fits Based on DBSCAN Clustering

Object-aware Semantic Mapping of Indoor Scenes Using Octomap

Indoor Interior Segmentation with Curved Surfaces Via Global Energy Optimization

VOX2BIM+ - A Fast and Robust Approach for Automated Indoor Point Cloud Segmentation and Building Model Generation