Automatic Estimation of Road Slopes and
Superelevations Using Point Clouds
Jin Wang, Zhenqi Hu, Yanyan Chen, and Zhiqing Zhang
Abstract
Road slopes and superelevations are important elements in
road widening and self-navigation investigations. This article
proposes to automatically extract road slopes and superele-
vations from data collected by a mobile laser scanning (
MLS
)
system. The proposed method is implemented as follows: (a) a
hierarchical segmentation is constructed to dynamically retain
point clouds on the road surface; (b) the point clouds are
divided into dynamically segmented blocks using trajectory
points, and plane estimation is investigated to continuously es-
timate road slopes and superelevations; and (c) the estimated
slopes and superelevations are iterated and filtered to decrease
the influence of outliers. The proposed method was tested and
validated for inclining and winding road types. The estimated
slopes and superelevations satisfied real road conditions.
Introduction
Mobile laser scanning (
MLS
) technology provides a new
approach for performing transportation characterization
research (e.g., El-Halawany
et al
., 2012; Holgado-Barco
et al
.,
2014; Vaaja
et al
., 2011). Road slopes and superelevations
play important roles in road inspection and maintenance.
Such characterizations also influence driver behavior in terms
of speed, acceleration, and lateral position. The availability of
detailed digital cartography for road slopes and supereleva-
tions is very important in topographic applications, such as
three-dimensional (3D) simulations and unmanned vehicles.
However, the process of automatically and precisely obtaining
road features from massive point clouds is challenging and
requires further investigation.
MLS
technology can collect accurate 3D geospatial data
with a high density at highway speeds for a lower cost than
traditional survey approaches. In the transportation disci-
pline,
MLS
does not require road closures or traffic disruption,
and synthesis applications have been illustrated by Williams
et al
. (2013). To some extent, contracting
MLS
for highway
surveying saves money in road closure fees and reduces costs.
Several
MLS
systems have appeared on the market due to the
advancement of laser scanning-related component tech-
nologies (e.g., scanning, imaging, and positioning devices)
(Graham, 2010). Although there are numerous companies
and research groups providing data processing services and
solutions for road asset inventory, management, and mainte-
nance (Gordon, 2010), the development of
MLS
software and
automated algorithms for extracting road features remains rel-
atively slow compared to the advancement of
MLS
hardware
(Yang
et al
., 2013).
MLS
-related technologies have been investigated to
extract road elements. Accurate cross-sections and digital
terrain models of an existing roadway have been provided
by fixed terrestrial laser scanners (Moscetti and Vespremi,
2002). This highly detailed information provides cross-sec-
tions at closer intervals, 3D elevations, and additional
location information for future applications. Roads extract-
ed from light detection and ranging (lidar) data have been
investigated using the method of classification, road value
processing and filling gaps in the road network (Matkan
et
al
., 2014). Lakakis
et al
. (2013) presented a mobile, low-
cost, dual differential global positioning system (
DGPS
) for
rapidly tracing basic road design elements. This system’s
accuracy was compared after its construction with classi-
cal surveying methods to verify its results. This research
demonstrated the advantages of using mobile global posi-
tioning systems (GPSs). One would expect an
MLS
system to
have greater advantages in computing road design elements
due to its rapid scanning and high-density point clouds.
In the field of roads and subsidiaries, researchers have
investigated extraction methods for road markings (Yu
et
al
., 2015a), road poles (EI-Halawany and Lichti, 2013), road
roughness (Kumar
et al
., 2015), road boundaries (Wang
et
al
., 2015), road horizontal alignment (Holgado-Barco and
González-Aguilera, 2015), road scene labeling and extract-
ing (Luo
et al
., 2016; Yu
et al
., 2016) and other street-scene
objects in three dimensions. Filtering off-road points is the
basis for extracting road surface elements. To estimate the
excavation volume for mountain road widening, Wang
et al
.
(2013) first processed the point cloud to a uniform grid and
separated road areas from off-road regions. Subsequently, the
excavation required to widen the road was determined using
segmentation. Yang
et al
. (2013) partitioned
MLS
point clouds
into a set of consecutive lines using
GPS
time measurements.
A moving window operator was used to filter out non-ground
points, and curb points were detected based on curb patterns.
The detected curb points were tracked and refined. If curb
points cannot be correctly detected, they may influence the
accuracy of subsequent analyses.
Based on generated images, road markings were extracted
with reflection information and segmented points (Yang
et
al
., 2012). Later, Guan
et al
. (2014) proposed a curb-based
extraction method to extract road surfaces. With the generated
intensity images, segmentation and morphological operations
were used to extract road markings. However, when estimating
vertical road elements, the generated images may become less
beneficial due to a lack of original 3D coordinate information.
In order to extract road related facilities, Euclidean dis-
tance clustering, segmentation and object matching frame-
work were developed to extract 3D objects (Yu
et al
., 2015b).
Jin Wang, Yanyan Chen, and Zhiqing Zhang are with the
Beijing Engineering Research Center of Urban Transportation
Operation Support Beijing University of Technology, 100124,
Beijing, China
)
Zhenqi Hu is with the Institute of Land Reclamation and Eco-
logical Restoration, China University of Mining and Technolo-
gy (Beijing), 100083, Beijing, China
.
Photogrammetric Engineering & Remote Sensing
Vol. 83, No. 3, March 2017, pp. 217–223.
0099-1112/17/217–223
© 2017 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.83.3.217
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
March 2017
217
