Space Classification from Point Clouds of Indoor Environments Based on Reconstructed Topology

Abstract

Reconstruction of as-built building models from point cloud data is a challenging problem with promising applications in the construction industry. In this paper, we outline the general concept of a data processing pipeline that produces fully three-dimensional, semantically rich and topologically valid as-built building models. Point cloud data is processed with a combination of histogram, voxel-based and RANSAC-based methods to detect surfaces of spaces and building components. Topological relations between building components (walls, slabs) are derived from a space partitioning that is generated from detected surfaces. The output from topology reconstruction is used as input for a space classification procedure which involves assigning functional properties to spaces. Each step in the data processing pipeline is illustrated with examples. Limitations of the proposed approach are discussed and an outlook of future development in this area is given.