IMAGE ORIENTATION BY
COMBINED BUNDLE ADJUSTMENT WITH FIXED IMAGERIES
Xiuxiao YUAN Chou XIE Shugen WANG
School of Remote Sensing and Information Engineering, Wuhan University
129 Luoyu Road, Wuhan, 430079, PR. CHINA
yxxqxhyw@public.wh.hb.en
Commission IT, WG II/1
KEY WORDS: Photogrammetry, DEM/DTM, Orthoimage, Exterior, Orientation, Adjustment, Accuracy
ABSTRACT
According to self-calibration bundle block adjustment, a mathematical model which solutes exterior orientation elements of the new
imageries by using fixed imageries in the same area is set up at first. The orientation parameters of the fixed imageries are previously
obtained. Several sets of aerial imageries at different scales are simulated based on a piece of DOM (Digital Orthophoto Map) with a
scale of 1:50,000 and corresponding DEM (Digital Elevation Model) with the interval of grid being 100 meters of some certain area.
Finally, the exterior orientation elements of these new imageries and 3D coordinates of photogrammetric points are calculated by
combined bundle block adjustment with the derived mathematical model. The empirical results show that the exterior orientation
elements and the coordinates of the photogrammetric points are identical between the mentioned method and the conventional bundle
block adjustment for the two periods of imageries at the same scale. However, when the image scales are different the adjusted
accuracy has an apparent decrease but can still meet the specification of topographic maps.
The experiment has verified that the determination of the orientation parameters of the new imageries by using the fixed imageries is
correct in theory and is feasible in practice. And the accuracy of 3D coordinates of the photogrammetric points, which are purely
gained by combined adjustment with two periods of imageries, satisfies the requirement of the specification of topographic maps.
That is to say, aerial triangulation without ground control points (GCPs) can be truly realized. It will have glory application future on
the map revision, the update of geographic spatial information database and automatic change detection of multi-temporal imageries.
1. INTRODUCTION
As is now well known, the key task for aerial photogrammetry
is to determine the orientation parameters of aerial imageries
quickly and correctly. For a long time, this goal has been
indirectly achieved by aerotriangulation with a lot of GCPs. In
order to reduce GCPs in photogrammetry, a combined block
adjustment for photogrammetric observations and various
auxiliary data was studied since the beginning of 1950s.
Despite many attempts, none of these efforts was able to
achieve a practical breakthrough because the used instruments
were very expensive and the auxiliary data had poorer quality.
In the 1970s, the USA developed GPS. It offers new
possibilities to extremely precise positioning of airborne
cameras and sensors during the photo flights. The technology
of combined block adjustment with 3D coordinates of the
camera stations, which determined by differential GPS
positioning based on carrier phase observations, avoids hard
field survey for GCPs and especially provides a topographic
map. of inaccessible regions of the Earth. This is called
GPS-supported aerotriangulation, which has led to a small
technological revolution in aerial photogrammetry (Ackermann,
1994). However, the superiority of GPS-supported
aerotriangulation only lies in photogrammetric procedure of the
great field, the medium and small scale, or inaccessible regions
of the Earth (Yuan, 2000). In the 1990s, combined use of GPS
and inertial measurement unit (IMU) was studied to determine
the orientation parameters during the photo flights in aerial
photography. According to current experiments, this combined
system of GPS/IMU is too expensive and the accuracy of
exterior orientation elements cannot satisfy the requirements of
photogrammetry at large and medium scale (Cannon, 1996;
Karsten, 2000; Gruen, 2001; Helge, 2002; Yastikli and
Jacobsen, 2002).
With the development and widely application of the digital
photogrammetry, a great count of 4D products (DEM, DOM,
DLG, DRG) has been produced and corresponding database
have been set up. In these products, DEM database records 3D
coordinates of ground objects, and DOM database stores the
orthoimages. These data can be conveniently accessed and used.
At the same time they need continual update. In addition, the
aerial photography with GPS navigation system can take photo
at an appointed position. It supplies fast technologies for
revision of maps and update of geographic spatial information
database. During these operations, if we can use some fixed
imageries to get orientation parameters of the new imageries
correctly, the aerial photogrammetry will become greatly
simplified. Based on this thought, this paper studies how to set
up a mathematical model which solutes exterior orientation
elements of the new imageries by using fixed imageries in the
same area, and how to determine the orientation parameters of
the new imageries and the positions of ground objects by
combined adjustment with two periods of imageries. The paper
aims at using existing data of DEM and DOM in the same area
and not using any GCP and data of GPS/IMU to determine
orientation parameters of new aerial imageries quickly. It tries
to verify the correctness and the feasibility of the theory
presented in the paper by simulated experiments, and to realize
really aerial triangulation without any GCP.
Intern
2
General!
adjustme
Where,
N e
Accordin
each poi
The empi
695km2
terrain is
aerial sim
camera at