You are using an outdated browser that does not fully support the intranda viewer.
As a result, some pages may not be displayed correctly.

We recommend you use one of the following browsers:

Full text

Proceedings International Workshop on Mobile Mapping Technology
Li, Rongxing

C. Tao, Assistant Professor
Dept, of Geomatics Engineering, The University of Calgary
2500 University Dr., NW, Calgary, Alberta, Canada, T2N 1N4
Tel: (403) 220-5826, Fax: (403) 284-1980, Email: ctao@uc;dgary.ca
KEY WORDS: Mobile mapping, transportation GIS, line feature extraction, multiple image matching, dynamic image sequences analysis,
and multinocular line reconstruction.
A new approach is developed for automated verification and updating of the transportation objects, such as traffic signs, street light poles, etc.,
in a transportation database using road image sequences collected by the mobile mapping system. The existing transportation databases are
used to predict the positions of the desired objects. Since the image sequences have been georeferenced using a GPS/INS technique, the
positions of these desired objects in the images can be determined. Multiple constraints available in the mobile imaging system are utilized, and
multiple image based feature extraction, image matching and line reconstruction algorithms are designed to detect objects in images and then
update an object database. The evaluation of the approach in terms of reliability and accuracy has been conducted using real image tests. The
results demonstrate that the road image sequences can be employed to update transportation databases in an automatic and cost-effective
Recently, GIS-T (transportation) has gained a great attention in both
GIS and transportation communities. The maintenance and updating
of transportation infrastructure databases along road corridors has
been a challenging issue to transportation departments. Are not only
the accuracy but also the currency of data stored in these database
critical to GIS-based planning, analysis and management. Preferably,
these databases need to be updated on a yearly or bi-yearly basis.
Figure 1. VISAT mobile mapping system
Mobile mapping technology has been proven an efficient and cost-
effective approach for collecting transportation related spatial data
along road corridors. The van-based mobile mapping system
integrating multiple sensors, GPS, INS and CCD cameras is able to
offer high-quality and georeferenced digital image sequences of road
scenes. The 3-D coordinated of any object of interest appearing in at
least two images can be determined by means of a post-mission
photogrammetric processing. At the University of Calgary, a
significant amount of research has devoted towards development of
multi-sensor integrated technology and employment of mobile
mapping images for various GIS applications. This research work
was conducted based on the VISAT mobile mapping system
(Schwarz et al., 1993). In the current implementation, eight CCD
cameras have been mounted on the top of a van so that the system is
able to capture road images with a field of view of 180 degree
(Figure 1).
In most urban areas, transportation databases already existed.
However, to update the databases is not only time-consuming but
also labor-intensive, since field surveys are extensively involved. It is
a fact that a survey crew has to be sent out to examine the conditions
of infrastructure, mark out missing objects, and re-survey moved
objects on a regular basis. In this paper, an innovative approach is
proposed to automate the procedure of examining and updating
transportation objects with linear features in a database using both
mobile mapping image sequences and existing object databases.
Most transportation objects have distinctive line features in images
and these line features are approximately vertical with respect to the
ground. This type of objects forms a large group in transportation
infrastructure, such as, traffic signs, street light poles, power-line
poles, and fire hydrants. Due to this reason, in the proposed
approach, a line feature extraction algorithm along with a
multinocular line matching algorithm are developed to detect and
verify the objects from image sequences, and a multinocular line
based photogrammetric reconstruction technique is then used to
determine the 3-D coordinates of the objects.
The literature regarding extraction of line features is very extensive
(Chen and Huang, 1990; Engelbrecht and Wahl, 1988; Liu et al.,
1990). The major advantage of the approach proposed in this paper
is that multiple constraints are developed and utilized in the design
and implementation of the approach. These constraints include (1)
the prior information of object positions obtained from an existing
database, (2) stereoscopic and sequential multiple image geometry of
objects from georeferenced image sequences, and (3) vertical line
features of objects in the images.
The systematic overview of the proposed approach is illustrated in
Figure 2. It consists of three main modules: