Share Email Print

Proceedings Paper

Automatic pole-like object modeling via 3D part-based analysis of point cloud
Author(s): Liu He; Haoxiang Yang; Yuchun Huang
Format Member Price Non-Member Price
PDF $17.00 $21.00

Paper Abstract

Pole-like objects, including trees, lampposts and traffic signs, are indispensable part of urban infrastructure. With the advance of vehicle-based laser scanning (VLS), massive point cloud of roadside urban areas becomes applied in 3D digital city modeling. Based on the property that different pole-like objects have various canopy parts and similar trunk parts, this paper proposed the 3D part-based shape analysis to robustly extract, identify and model the pole-like objects. The proposed method includes: 3D clustering and recognition of trunks, voxel growing and part-based 3D modeling. After preprocessing, the trunk center is identified as the point that has local density peak and the largest minimum inter-cluster distance. Starting from the trunk centers, the remaining points are iteratively clustered to the same centers of their nearest point with higher density. To eliminate the noisy points, cluster border is refined by trimming boundary outliers. Then, candidate trunks are extracted based on the clustering results in three orthogonal planes by shape analysis. Voxel growing obtains the completed pole-like objects regardless of overlaying. Finally, entire trunk, branch and crown part are analyzed to obtain seven feature parameters. These parameters are utilized to model three parts respectively and get signal part-assembled 3D model. The proposed method is tested using the VLS-based point cloud of Wuhan University, China. The point cloud includes many kinds of trees, lampposts and other pole-like posters under different occlusions and overlaying. Experimental results show that the proposed method can extract the exact attributes and model the roadside pole-like objects efficiently.

Paper Details

Date Published: 26 October 2016
PDF: 16 pages
Proc. SPIE 10008, Remote Sensing Technologies and Applications in Urban Environments, 100080T (26 October 2016); doi: 10.1117/12.2242050
Show Author Affiliations
Liu He, Wuhan Univ. (China)
Haoxiang Yang, Wuhan Univ. (China)
Yuchun Huang, Wuhan Univ. (China)

Published in SPIE Proceedings Vol. 10008:
Remote Sensing Technologies and Applications in Urban Environments
Thilo Erbertseder; Thomas Esch; Nektarios Chrysoulakis, Editor(s)

© SPIE. Terms of Use
Back to Top
Sign in to read the full article
Create a free SPIE account to get access to
premium articles and original research
Forgot your username?