Structure-preserving Planar Simplification for Indoor Environments

TL;DR

The paper tackles structure-preserving indoor scene reconstruction by separating structured (walls-ceiling-floor) from non-structured content and applying a two-step approach: RANSAC-based planar primitive extraction followed by vertex translation and mesh clipping to preserve geometry, with surface reconstruction for non-structured parts. The method aligns data to the $Z$-axis under the Manhattan World, segments planes into ceiling/floor/walls, and generates simplified planar meshes that remain faithful to the scene layout, even for multi-story or slanted environments. Quantitative comparisons against PolyFit, KSR, and other shape-approximation and floorplan methods show reduced mesh complexity (fewer faces) and competitive RMSE, while qualitative results demonstrate robust handling of interior partitions and partial rooms. The work offers practical implications for efficient, structure-aware 3D indoor modeling and floorplan estimation, with potential applications in AR/VR, architectural design, and real estate visualization.

Abstract

This paper presents a novel approach for structure-preserving planar simplification of indoor scene point clouds for both simulated and real-world environments. Initially, the scene point cloud undergoes preprocessing steps, including noise reduction and Manhattan world alignment, to ensure robustness and coherence in subsequent analyses. We segment each captured scene into structured (walls-ceiling-floor) and non-structured (indoor objects) scenes. Leveraging a RANSAC algorithm, we extract primitive planes from the input point cloud, facilitating the segmentation and simplification of the structured scene. The best-fitting wall meshes are then generated from the primitives, followed by adjacent mesh merging with the vertex-translation algorithm which preserves the mesh layout. To accurately represent ceilings and floors, we employ the mesh clipping algorithm which clips the ceiling and floor meshes with respect to wall normals. In the case of indoor scenes, we apply a surface reconstruction technique to enhance the fidelity. This paper focuses on the intricate steps of the proposed scene simplification methodology, addressing complex scenarios such as multi-story and slanted walls and ceilings. We also conduct qualitative and quantitative performance comparisons against popular surface reconstruction, shape approximation, and floorplan generation approaches.

Figures (16)

• Figure 1: Overall system block diagram of our approach.
• Figure 2: Overview of our approach: (a) input point cloud, (b) planar primitives extraction (as vertex groups), scene segmentation into (c) structured and (d) non-structured scenes, (e) generation of simplified structured mesh, (f) surface reconstruction of non-structured scenes, (g) final scene mesh (with (h) its ceiling).
• Figure 3: Axis Alignment. (a) Unaligned mesh and (b) Axis-aligned mesh
• Figure 4: Vertex translation. The vertex $\mathbf{V}_1$ nearest to the line of intersection between $P_i$ and $P_j$ is translated to point $\mathbf{Q}_1$. The same process is repeated for vertex $\mathbf{V}_2$. $Z=0$ is an arbitrary plane to determine the point of intersection X.
• Figure 5: Mesh Clipping. (a) two rooms with intersecting wall meshes, $M_{P1}$ and $M_{P2}$, (b) ceiling plane, $M_{ceiling}$ with points clouds as dots, (c) $M_{P1}$ clipping $M_{ceiling}$ into planes a and b, (d)(e) $M_{P2}$ clipping the plane a into planes a$_1$ and a$_2$, (f) $M_{P2}$ clipping the plane b into planes b$_1$ and b$_2$, (g) final $\mathcal{M}_{ceiling}$. The normal vectors $\pm\hat{n}$ represent the directions of clipping.