Shape Grammar Approach to 3D Modeling of Indoor Environments Using Point Clouds

Abstract

As-is three-dimensional (3D) models of indoor environments are of paramount importance for a variety of applications such as navigation assistance and emergency response. Whereas manual reconstruction of 3D indoor models is a time-consuming task, methods for automatic or semiautomatic reconstruction can achieve a significant reduction of the time and labor required for indoor modeling. This paper proposes a new shape grammar approach for efficient generation of 3D models of indoor environments from point clouds. The shape grammar derives 3D models of complex environments by using a simple primitive and iterative application of grammar rules governed by a production procedure. The method can reconstruct both building elements and navigable spaces along with their topological relations. It produces 3D parametric models with high geometric accuracy and rich semantic content compliant with building information modeling (BIM) standards. Experiments on synthetic and real data sets show the ability of the method to generate highly accurate indoor models with computation times ranging from seconds to a few minutes. The usefulness of the output models in supporting indoor path planning at varying granularities is also demonstrated.