Hierarchical Space Partition for Surface Reconstruction
2026-06-03 • Computer Vision and Pattern Recognition
Computer Vision and Pattern RecognitionComputational Geometry
AI summaryⓘ
The authors tackle the challenge of turning messy 3D point cloud data from LiDAR scans into simple, accurate polygon shapes. They notice that some parts of a scene are missing because of scan limitations, so they develop a method that guesses and fills in these missing parts using a smart way to grow and organize flat surfaces (planes). Their approach classifies planes by how visible they are and grows them in order of importance to build a complete 3D model. They use a special optimization to create a clean, closed mesh from this, and tests show their method works better than others.
point cloudLiDARpolygonal meshplane extractionscene reconstructionhierarchical partitionmin-cut optimization3D visionocclusionsurface modeling
Authors
Minjie Tang, Xiangfei Li
Abstract
Generating compact polygonal models from point clouds is a key problem in 3D vision and computer graphics. However, due to inherent limitations of LiDAR scanning (e.g. range constraints and occlusions), critical scene information is often missing, leading to degraded reconstruction accuracy. To address this, we propose a plane assembling strategy that effectively recovers missing details while maintaining model compactness. We classify all the planes extracted from the scene into three categories: highly visible, barely visible, and invisible. The invisible planes, which are recovered by scene structure analysis, indicate the missing details. The three types of planes correspond to the three growth priorities. Each plane grows according to the priority level, and the space is partitioned progressively, namely, the hierarchical partition. Subsequently, we generate a watertight polygonal mesh from the partition via a min-cut-based optimization. Finally, comparisons on public datasets show the effectiveness and superiority of our method against mainstream approaches. The project page is available at https://hsr-3dv.github.io/.