lysis
age
vey.
onal
Ww in
ling
and
asel,
lage
ON
s of
and
s of
tion
Huijing Zhao
ROBUSTLY REGISTERING A NETWORK OF RANGE IMAGES OF URBAN OBJECTS
Huijing ZHAO , Ryosuke SHIBASAKI
Center for Spatial Information Science, Univ. of Tokyo, JAPAN
{chou,shiba}@skl.iis.u-tokyo.ac.jp
KEY WORDS: laser range image, registration, urban 3D object.
ABSTRACT
"There is a potentially strong demand for detailed 3D spatial data of urban area. Ground-based laser range
scanner is one of the promising devices to acquire range images of urban 3D objects. In this paper, the authors
propose a method of automatically registering a network of range images into a well-balanced model for the
reconstruction of 3D urban objects. Assuming relative transformation between neighboring range images have
been estimated using a pair-wise registration method, multiple registration is achieved in two steps, sequentially
aligning all range images into a global coordinate system and minimizing violation, which is caused by the
accumulation of pair-wise registration error. Experimental results are presented, where 42 range images are
registered to construct a 3D model of the buildings in the campus of University of Tokyo. Two sets of ground
truth, a 1:500 scale digital map and the location of viewpoints measured by GPS, are used to examine the
accuracy in multiple registration. Experiments are conducted in two levels. The first experiment examines
registration accuracy when location and viewing direction of range images are unknown, while in the second
experiment, assuming that several viewpoints has been measured using GPS, improvement to registration
accuracy using GPS points is studied. Through the experiments, it is demonstrated that the method has
accuracy in urban outdoor environment.
1 INTRODUCTION
In a variety of applications ranging from visualization of urban landscape to advanced automobile/pedestrian
guidance systems for ITS (intelligent transportation system), accurate and detailed 3D urban spatial database
is becoming strongly demanded. There are two approaches of 3D data acquisition; air-based and ground-based.
With air-based data acquisition techniques, typically, aerial survey, can cover relatively wide area, but may fail
to capture details of urban objects such as side walls (facade) of buildings. On the other hand, ground-based
method such as vehicle-borne CCD camera can easily cover such details of urban objects, though the spatial
coverage may be limited. Recently, reconstructing 3D urban objects using ground-based acquisition techniques
are attracting more attentions because such details of urban objects which can be easily viewed from streets or
on ground surface are found important for guidance system and radio disturbance analysis in telecommunication
and so forth. In many applications of 3D GIS, viewpoints of users are on the street or ground surfaces, not in
the air.
Several researches using ground-based CCD camera have demonstrated that 3D urban object can be extracted
using motion and stereo vision technique (e.g. S.Ozawa et.al.1998). However, insufficient robustness in stereo
matching, distortion from limited resolution and unstable geometry of CCD cameras are major obstacles to the
operational uses of these method. On the other hand, range scanners are also used for 3D object modeling.
Efficiency and accuracy of range scanner serving for spatial data acquisition has been demonstrated in Thorpe
et.al.1988 in a mobile navigation system; Kamgar-Parsi et.al.1991 in obtaining a map of ocean floor; Chen
and Medioni 1992, Champleboux et.al.1992, Shum et.al.1994 in modeling small objects such as teeth, status,
mechanical parts, etc.; Ng et.al.1998 in constructing 3D models of indoor objects; Lemmens et al. 1997, Haala
et al.1998 in air-borne systems. In recent years, with the development of eye-safe laser range scanner, ground-
based measurement using laser range scanner in urban environment becomes technically feasible. However,
there is still no other research addressed on the 3D urban outdoor object modeling using ground-based laser
range scanner. Since one snapshot can not cover the entire 3D object, one of the major challenges in applying
ground-based laser range scanner to the reconstruction of 3D urban objects is the registration of multiple
overlapping range images acquired at different locations (viewpoint) with different viewing angles. Registering
multiple range images - i.e., correctly aligning range images by transforming them into a global (common)
coordinate system - is typically solved as a two-step procedure, pair-wise registration and multiple registration.
In pair-wise registration, neighboring range images, which has a degree of overlay, are registered to find the
relative transformation between the coordinate system of two range images. In multiple registration, absolute
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000. 1033