Semantic Decomposition and Reconstruction
of Residential Scenes from LiDAR Data


Hui Lin1*, Jizhou Gao1*, Yu Zhou2, Guiliang Lu2, Mao Ye1, Chenxi Zhang1, Ligang Liu3 and Ruigang Yang1
1University of Kentucky, 2Nanjing University, 3University of Science and Technology of China
(*Joint first authors)
ACM Transactions on Graphics (SIGGRAPH 2013), 32(4)


Figure 1

Figure 1: Our reconstruction pipeline. From left to right: (a) semantically labeled 3D point cloud; (b) reconstructed objects using category-specific methods, including billboard trees, replaced common objects, and a building. The color-code on the building shows recognized different building parts; (c) textured 3D models on a ground plane, and (d) an overview of an automatically reconstructed large-scale scene


We present a complete system to semantically decompose and reconstruct 3D models from point clouds. Different than previous urban modeling approaches, our system is designed for residential scenes, which consist of mainly low-rise buildings that do not exhibit the regularity and repetitiveness as high-rise buildings in downtown areas. Our system first automatically labels the input into distinctive categories using supervised learning techniques. Based on the semantic labels, objects in different categories are reconstructed with domain-specific knowledge. In particular, we present a novel building modeling scheme that aims to decompose and fit the building point cloud into basic blocks that are block-wise symmetric and convex. This building representation and its reconstruction algorithm are flexible, efficient, and robust to missing data. We demonstrate the effectiveness of our system on various datasets and compare our building modeling scheme with other state-of-the-art reconstruction algorithms to show its advantage in terms of both quality and speed.


High Res (17Mb) – 10 pages PDF download
Low Res (1Mb) – 10 pages PDF download
Supplementary Material download
Bibtex download



60fps high resolution link



Coming Soon



Fast Forward





Figure 2

Figure 2: Hierarchy-tree representation of buildings.

Figure 3

Figure 3: Decomposition and Reconstruction Algorithm. From left to right, (a) input labeled point cloud (red as walls, blue as roofs) with missing data; (b) plane extraction; (c) block decomposition; and (d) complete reconstruction result. Each color represents an individual block.

Figure 4

Figure 4: Results on ground-based LiDAR datasets. Each row shows the result of one house. 1st and 2nd columns: point clouds color coding based on height in two perspectives, 3rd and 5th columns: reconstructed models, and 4th and 6th columns: reconstructed models with textures.

Figure 5

Figure 5: Qualitative evaluation on three building reconstruction algorithms. Four groups of comparison results are shown. In each group, from left to right: point clouds, results by Piecewise Planar Reconstruction, results by 2.5D Dual Contouring, and results by our approach.

FIgure 6

Figure 6: Quantitative evaluation on three building reconstruction algorithms. From left to right, results by Piecewise Planar Reconstruction, results by 2.5D Dual Contouring and results by our method. Top: point clouds color-coding based on point to surface distance. Bottom: the corresponding reconstruction results.

Figure 7

Figure 7: (left) Results on Wright-State-100 airborne dataset; (right) an overall view of reconstructed houses in an upscale residential area.


We would first like to thank the anonymous reviewers for their valuable feedback. Also, we would especially like to thank Yongwook Song for LiDAR data acquisition and processing, Qing Zhang, Aaron Camenisch, Seth Parker for video editing, Wenmeng Zhou and Yiping Yang for data labeling. This work is supported in part by US NSF grant IIS-0448185, CCF-0811647, and CNS-0923131. Yu Zhou is supported by the National Natural Science Foundation of China (61100111). Ligang Liu is supported by the National Natural Science Foundation of China (61222206) and the National Basic Research Program of China (2011CB302400).



@article {lin_sig13,

title = {Semantic Decomposition and Reconstruction of Residential Scenes from LiDAR Data},

author = {Lin, Hui and Gao, Jizhou and Zhou, Yu and Lu, Guiliang and Ye, Mao and Zhang, Chenxi and Liu, Ligang and Yang, Ruigang},

journal = {ACM Transactions on Graphics, (Proc. of SIGGRAPH 2013)},

volume = {32},

number = {4},

year = {2013},