RAPID UPDATE OF ROAD SURFACE DATABASES USING MOBILE
LIDAR: ROAD-MARKINGS
Haiyan Guan
a
, Jonathan Li
b, a,
*, Yongtao Yu
b
, Cheng Wang
b
a
Dept. of Geography and Environmental Management, University of Waterloo, 200 University Avenue West, Waterloo, Ontario
N2L3G1, Canada -h6guan@uwaterloo.ca; junli@uwaterloo.ca
b
Key Laboratory of Underwater Acoustic Communication and Marine Information Technology (MOE), Xiamen University, Xiamen, FJ
361005, China- junli@xmu.edu.cn; allennessy.yu@gmail.com; cwang@xmu.edu.com,
Abstract
Road surface markings are used on paved roadways to provide guidance and information to drivers and
pedestrians, which are a critical feature in the traffic management systems. This paper presents an automated
approach to detection and extraction of road markings from mobile laser scanning (MLS) point clouds by taking
advantages of multiple data features. To improve computational efficiency, the raw MLS point cloud data are first
converted to geo-referenced images, based on elevation, intensity and point density, using inverse distance
weighted interpolation, respectively. Afterwards, three filters are designed to extract road markings step-by-
step: (1) the elevation filter is used to generate an elevation mask to remove high objects from the geo-referenced
intensity image; (2) the point density filter is implemented to extract road surfaces in the geo-referenced intensity
image; (3) the filtered geo-referenced intensity image is processed by thresholding and point density to obtain
road markings, followed by a Canny detector and Hough transform used to extract straight-lines of road
markings. Two RIEGL VMX-450 datasets demonstrate that the proposed multi-feature road marking extraction
method has a good performance of road marking extraction from a large volume of mobile laser scanning data.
Keywords: MLS, Road Marking, Geo-Referenced, Point Density, Intensity, Elevation
Introduction
Road surface markings are used on paved roadways to provide guidance and information to drivers and pedestrians,
which are a critical feature in the traffic management systems. For example, driver assistance systems require a
reliable environment perception for the improvement of the traffic safety by informing the motorists and preventing
accidents. In line with the condition of the pavement, the topography of the road, the visibility of road markings are the
key elements in accidents where the road itself is the cause. Especially, in the urban environments, with the increase
of population and urbanization, high accident rates are caused by the absence of clearly-presented road signalisation
[1]. Thus, road infrastructure divisions need a practical tool that can monitor the situation of road markings to maintain
to high technique standards for perfect visibility of road markings, accordingly.
McCall and Trivedi summarized lane marking detection techniques that include the use of edges, regions and tracking
for continuity [2]. Li et al. presented a method, which uses a fuzzy-reasoning-based general technique for edge
detection to classify a pixel into a uniform region based on luminance differences between the pixel and its neighbors
[3]. Li et al. detected arrow markings on road surfaces based on shape information [4]. Most algorithms of road
markings recognition are commonly comprised by the extraction of candidates and a classification step [5-9].
Although works on the detection of road markings from either digital photographs or videos have been pursued for a
number of years, the image based recognition systems are limited to extract precise geometry information under poor
illumination conditions. Compared to photogrammetry, laser scanning, as an active remote sensing technology,
captures very highly accurate 3D point clouds with a high point density in a relatively short amount of time [10-12].
Typically, a mobile lidar technology is ideally suited for corridor mapping due to its “drive -by” data acquisition
pattern, and therefore market focus is on road and rail networks. This technology collects survey-grade data of
* Corresponding author: junli@xmu.edu.cn, junli@uwaterloo.ca.
2013 Fifth International Conference on Geo-Information Technologies for Natural Disaster Management
978-1-4799-2269-7/13 $26.00 © 2013 IEEE
DOI 10.1109/GIT4NDM.2013.22
124
2013 Fifth International Conference on Geo-Information Technologies for Natural Disaster Management
978-1-4799-2269-7/13 $26.00 © 2013 IEEE
DOI 10.1109/GIT4NDM.2013.22
124