In: Stilla U, Rottensteiner F, Paparoditis N (Eds) CMRT09. IAPRS, Voi. XXXVIII, Part 3/W4 — Paris, France, 3-4 September, 2009
SURFACE MODELLING FOR ROAD NETWORKS USING
MULTI-SOURCE GEODATA
Chao-Yuan Lo , Liang-Chien Chen, Chieh-Tsung Chen, and Jia-Xun Chen
Department of Civil Engineering, National Central University, Jhungli, Taoyuan 32001, Taiwan -
freezer@csrsr.ncu.edu.tw
Center for Space and Remote Sensing Research, National Central University, Jhungli, Taoyuan 32001, Taiwan -
lcchen@csrsr.ncu.edu.tw
Department of Land Administration, Taipei 10055, Taiwan -
{moi5383; moil240}@moi.gov.tw
Commission III, WG III/4
KEY WORDS: Surface, Reconstruction, Three-dimensional, Geometric, Laser scanning, Modelling
ABSTRACT:
Road systems are the fundamental component in the geographic information systems. This kind of civil infrastructures has large
coverage and complex geometry. Thus, the modelling process leads to handling huge data volume and multi-source datasets. A
reasonable process should be able to reconstruct separate parts of road networks and combine the surfaces together. Hence, the
reconstruction of complete three-dimensional road networks needs scrutiny when a large area is to be processed. This paper proposes
a scheme to focus on this issue using an integrated strategy with multi-source datasets. The modelling processes combine different
data sources to refine road surfaces to keep the continuities in elevation and slope. The proposed scheme contains three parts: (1) data
pre-processes, (2) planimetric networking, and (3) surface modelling. In the first part, datasets are registered in the same coordinate
system. In the next step, topographic maps provide the roadsides to derive the geometric topology of road networks. Finally, those
centerlines combine airborne laser scanning data to derive road surfaces. Considering the data variety, some road segments generated
from aerial images are also included in the proposed scheme. Then, the successive process integrates those models for the refinement
of road surfaces. The test area is located in Taipei city of Taiwan. The road systems contain local streets, arterial streets, expressways,
and mass rapid transits. Some roadways are multi-layer and cross over with different heights. The final results use three-dimensional
polylines and ribbons to represent geometric directions and road surfaces. Experimental results indicate that the proposed scheme
may reach high fidelity.
1. INTRODUCTION
Based on the viewpoint of decision support for modem cities,
the reconstruction of a virtual environment is an essential task.
The applications include urban planning, traffic simulation, true
orthorectification (Zhou et al., 2005), hazard simulation,
communication, etc. Since the road models are one of the most
prominent components in the urban information systems, the
reconstruction of the model becomes increasingly important. In
general, the traditional topographic map is a kind of widely used
dataset that describes road geometries. It can efficiently build
single-layer road models. However, this civil infrastructure is
developed rapidly in modem cities for the traffic demand, and
road types become more complex including local streets,
arterial streets, expressways, freeways, and mass rapid transit.
Single-layer road networks have changed to multi-layer systems
and topomaps may be insufficient to describe complex roads.
The elevation information of road surfaces needs to be
considered for the separation of overpasses.
Some researches focused on the surface modelling processes
with different strategies and data, e.g. aerial photos, laser
scanning data, GPS data, topomaps, and so on. Cannon (1992)
proposed a scheme to locate the three-dimensional road profiles
integrating GPS and INS data. A related work also had been
made to estimate the slope information of road profiles using
GPS data (Han and Rizos, 1999). Some studies preferred to
derive road information in spectral domain. They analyzed road
shapes of centerlines or boundaries to derive road geometries
with vehicle-based images (Yan et cl., 2008), aerial photos
(Treash and Amaratunga, 2000; Hinz and Baumgartner, 2003;
Dal Poz et al., 2004), satellite images (Yan and Zhao, 2003;
Doucette et al., 2004; Hu et al., 2004a; Kim et al., 2004; Karimi
and Liu, 2004; Yang and Wang, 2007), airborne laser scanning
data (Clode et al., 2007). Some proposed semi-automatic
approaches basing on the matching technique to reliably extract
road geometries with manual editing from high-resolution
satellite imagery (Hu et al., 2004a; Kim et al.,2004). Easa et al.
(2007) focused on the automatic image processing to extract
edge lines for calculation of geometric parameters to describe
horizontal alignments from high resolution images.
On the other hand, an integrating strategy had been proposed to
deal with this issue using aerial images and laser scanning data
(Hu et al., 2004b; Zhu et al., 2004). Zhang (2003) integrated
aerial photos and geo-database to derive and update three-
dimensional road data. Moreover, geo-database and laser
scanning data also could be a combination. Hatger and Brenner
(2003) calculated the profile geometries of centerlines from the
geo-database and digital surface models. The segment-based
method used region growing to detect road areas for the
calculation of geometric parameters to refine the geo-database.
Furthermore, Cai and Rasdorf (2008) also combined two
datasets, airborne laser scanning data and planimetric centerline
Corresponding author
