SPATIAL LOCATION ON CITY 3D MODELING WITH CLOSE-RANGE STEREO
IMAGES
Jia Sheng-ju®*, Chen Ying-ying *, Li Ying-cheng"
e : eJ o e >
“ Remote Sensing Research Center of Tongji University Shanghai China 200092 shengju_jia@sina.com
? Chinese Academic of Mapping and Surveying Beijing China 100039
Commission V, WG V/2
KEY WORDS: Close-range, Free network, Bundle adjustment, Relative control
ABSTRACT:
This paper chooses close-range target stereo image data as main topic of 3DCM, and carries through comprehensive and systemic
study on photogrammetry of free network bundle adjustment from theory to practice. On the data processing of spatial model
location, free network bundle adjustment is introduced to solve the base of spatial location. Photogrammetry and non-
photogrammetry information are integrated to develop a union adjustment system with self-calibration parameter. Simultaneously
solve problems of the big angle linked of different models that take photo around a building, and none-control point and only
relation controls used to carry out free network adjustment. Process of realization and results are given.
1. INTRODUCTION
The 3DCM(three dimension city modeling) is hotspot of digital
photogrammetry, computer vision, GIS and other correlation
study. The basic geography data and methods are utilized by
model create tools such as GIS systems, image processing to
create three dimensions model. Measures of target acquiring
include photography, laser scanning and video-theodolite, etc.
Highly effective works are presented in spatial 3D modeling.
But 3D create tools are different from automatic and half-
automatic modeling in GIS field. Remote sensing images
applied in three dimensions modeling are effective measures.
But this measure cannot solve shelter from high building, the
side face, complex construction, and special working face of
building. Cannot satisfy high scale, accuracy (cm) necessity of
city and engineering. As technology supplement of remote
sensing, there are significance researching on methods of high
speed data acquiring and processing to close-range target, and
realizing real-time, automatic, and modeling to closed-range
objects.
Acquiring of threc dimensions spatial information of close-
range object include fusion of multi-source informs, correlation
matching of image, and processing of digital photogrammetry
etc. On the basis of reviewing the current research status of
3DCM , based on multi-resource information, this paper
chooses close-range target sterco image data as main topic of
3DCM, and carries through comprehensive and systemic study
on photogrammetry of free network bundle adjustment from
theory to practice. On the data processing of spatial model
location, classic network and especially free network bundle
adjustment is introduced to solve the base of spatial location.
Photogrammetry and non-photogrammetry information are
integrated to develop a union adjustment system with self-
calibration parameter. Simultaneously solve problems of the big
angle linked of different models that take photo around a
building, and none-control point and only relation controls used
to carry out free network adjustment.
2. CONSTRUCT CLOSE-RANGE FREE NETWORK
In close-range photogrammtry, absolute orientation of the target
objects with respect to control points does not play a significant
role. What is important is to measure the relative positions of
points on surface of an object, and determine its size, shape and
volume according to desired accuracy, especially on 3d
modeling in city close-range photogrammtry. Three dimension
objects reconstructed and recovered from images, inner oriented
data and relate control condition need mathematic models and
algorithms to realize multi-resource spatial data integration
under universal coordinate system.
2.1 Coordinate System Select
The independent image pair coordinate system is selected as
basic coordinate system. SI, S2 being the projection centres,
and the direction of the baseline B is taken as the X axis of a
temporary coordinate system, and the direction of the Y axis is
perpendicular to the principle epipolar plane of the left-hand
photograph (see Fig 2-1 image pair). The elements of inner
oricntation, focal, and pixel size of image can be acquired
through calibration in laboratory. Automatic matching or
semiautomatic measure can acquire coordinates of image points
of photo pair.
2.2 Solution Model Coordinates
In the case of the rotation system in which the Y axis is taken as
the primary axis, the elements of relative orientation of the
photo pair are 7, , K, (left photo), €, 7,,Æ, (right photo).
When SI is taken as the origin, the locations of S1 and al with
respect to ST can be determined from the vectors SS and
Sal (see Fig. 2-1), whereas the location of image point a2
with respect to S2 can be determined from the vector S
Since these three vectors are in the same epipolar plane, they
are coplanar, which means their scalar-vector product of is zero:
5,5, e(5,a, x 5,a,) -0 (D
Which, express in coordinates, is:
Y «dc
! , =0 (2)
yy
* Jia sheng-ju, PhD. Study on digital close-range photogrammetry and 3d visualization.
Inter
(À,
COOT(
2.3
The
base!
pair
adjus
and
coor
After
point
carri
in th
is to
first |
of th
range
inder
The
analc
wher
rotati
mode
unde:
mode
whol
