Indoor 3D Point Cloud Segmentation Based on Multi-Constraint Graph Clustering

Indoor scene point cloud segmentation plays an essential role in 3D reconstruction and classification. This paper proposes a multi-constraint graph clustering method (MCGC) for indoor segmentation. The MCGC considers multi-constraints, including extracted structural planes, local surface convexity, color information of objects Firstly, the raw is partitioned into patches, we propose robust plane extraction to extract main planes scene. Then, match between patches achieved by global energy optimization. Next, closely integrate multiple constraints mentioned above design algorithm partition cluttered scenes object parts. Finally, present post-refinement step filter outliers. We conducted experiments on benchmark RGB-D dataset real laser-scanned perform numerous qualitative quantitative evaluation experiments, results which have verified effectiveness method. Compared with state-of-the-art methods, can deal more efficiently restore details structures. segment precision recall experimental reach 70% average. In addition, great advantage that speed processing clouds very fast; it takes about 1.38 s data 1 million points. It significantly reduces computation overhead achieves real-time