ANNER 
TROL 
finimum GCP, Platform 
system developed by 
y linear ground objects 
near imaging sensors, 
;h obtained image line 
: errors such as offsets 
inate system axes, and 
achieve high pointing 
| orientation strategies 
he real flight data and 
| be valuable for other 
s with a comparable size 
le presently and are not 
example, SPOT 1, 2, and 
lect stereo imagery using 
MOMS-02, IKONOS-2, 
> along-track technique to 
  
> —— 
flight direction 
      
forward 
cteristics of TLS. 
  
nd inertial | (GPS/INS) 
mmercial airborne digital 
into practical applications 
\ (Digital Photogrammetry 
ted by the Institute of 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004 
  
Photogrammetry, University of Stuttgart to produce 1:25,000 
and 1:50,000 scales maps and automated DTM generation to 
accuracies of better than 3m in 1995. The LH Systems ADS40 
airborne digital sensor was developed by LH Systems Co. Ltd in 
2000 and has been put on the world market. STARLABO 
Cooperation designed the helicopter-mounted high resolution 
TLS imaging system STARIMAGER jointly with University of 
Tokyo in 2000 and have finished several test flights and 
practical applications. These imaging systems use three linear 
arrays mounted in the sensor focal plane to collect forward-, 
nadir-, and backward-looking imagery for stereo mapping. The 
concept of three linear scanners to collect stereo imagery has 
been described in many literatures (Chen, 2001; ) and is 
illustrated in Figure 1. 
Unlike frame photography, where all pixels in the image are 
exposed simultaneously, each line of TLS image is collected in a 
pushbroom fashion at a different instant of time. Therefore, there 
is in principle a different set of values for the six exterior 
orientation elements for each line of the pushbroom scan. 
Although the traditional indirect approach using ground control 
points for the determination of the exterior orientation elements 
of frame photographs through standard aerial triangulation 
works for frame aerial photographs and spaceborne imagery, 
this process if highly inefficient. This is because satellite 
platforms remain relatively stable in relation to their orbital 
altitude; any deviation of attitude from normal is usually minor 
and systematically spread over the entire satellite scene 
(Christensen et al., 1988), so mathematical sensor models are 
developed to recover the time-dependent position and 
orientation of the scanner. Airborne scanners on the other hand 
are subject to atmospheric turbulence during their flight that can 
lead to severe image distortions in the raw TLS imagery. For 
airborne TLS a direct processing strategy utilizing direct 
measurements of the exterior orientation provided by GPS and 
INS is necessary for operational and efficient data evaluation. 
Even though direct georeferencing is no must for digital frame 
cameras a GPS/INS component is also included in some systems 
(Toth, 1998). 
The purpose of this paper is to deal with the integration of GPS, 
INS, and STARIMAGER imagery for the georeferencing of a 
digital airborne linear camera system with minimum ground 
control. In this paper we tested STARIMAGER imagery with a 
block of six strips and different number and geometric 
configuration of ground control points and reported our obtained 
results which could be taken as theorical reference for practical 
TLS imagery triangulation and other linear imaging system 
imagery geo-referencing process. 
2 Combined Bundle Adjustment with GPS/IMU for 
STARIMAGER 
The STARIMAGER is equipped with a GPS/IMU system to 
record the position and attitude of each image line during the 
flight. However, like other frame sensor equipped with 
GPS/IMU, the use of GPS and IMU for line sensor also requires 
that certain measures be taken before and after the flight because 
the positions and orientations from GPS/IMU do not refer to the 
perspective center of the imaging sensor directly. Caused by 
translational and rotational offsets, the GPS antenna and the 
center of the inertial system are displaced from the camera. 
Additionally, the attitudes from GPS/IMU are calculated from 
the rotation of the IMU body frame defined by the IMU sensor 
axes to the local level frame. The IMU axes do not coincide with 
the image coordinate frame. The translational offsets between 
GPS antenna and perspective center of camera can be 
determined using conventional terrestrial surveying methods 
after installation of the system in the aircraft used for the 
imaging flight. The rotational offsets between the IMU sensor 
axes and the camera coordinate system cannot be observed via 
conventional survey methods. Therefore, these rotational offset 
or misalignment angles between the IMU and camera system 
have to be determined with triangulation using a small number 
of tie and control points similar to conventional aerial frame 
camera. In addition to these offsets and misalignments, some 
systematic errors from GPS/IMU such as drifts of IMU should 
be considered in triangulation. The primary focus of this section 
is to present mathematical models used in triangulation of 
STARIMAGER to deal with the systematic errors from 
GPS/IMU observing data. Therefore, we will first describe the 
error sources in GPS/IMU and then give two algorithms to 
remove the systematic offsets for obtaining accurate exterior 
orientation parameters for STARIMAGER imagery. 
2.1 GPS/IMU data process 
The GPS/IMU data process is an important step towards high 
quality imagery and accurate measurements derived from it. The 
timing of IMU recording, GPS recording and CCD line 
recording must be done using a synchronized clock. This allows 
the precise registration of each data-recording event. Due to 
different sampling frequencies of GPS/IMU, a special software 
was developed to post-process their original data including GSP 
data re-sampling according to image line record and coordinate 
system conversion. 
As most tests and applications by integrating GPS/IMU systems 
for geo-reference of image data, the positions and attitudes from 
GPS/IMU do not refer to the perspective center of the imaging 
sensor directly (Chen ef al., 2001; Lee et al., 2000). Caused by 
translation and rotation offsets, the GPS antenna and the center 
of the IMU are displaced from the camera. Additionally, the 
attitudes from GPS/IMU are calculated from the rotation of the 
IMU body frame defined by the IMU sensor axes to the local 
level frame. The IMU axes do not coincide with the image 
coordinate frame. These offsets have to be taken into account 
before applying the orientations for the georeferencing of the 
imagery. The translation offsets are determined using 
conversional terrestrial surveying methods after installation of 
the system in the satellite and aircraft used for the photo flights. 
The rotation alignments between IMU and camera coordinate 
system cannot be observed via conventional surveying methods. 
Additionally, there are some drift errors caused by remaining 
sensor offsets. Therefore, these alignments and drift errors have 
to be determined with in-flight calibration using a small number 
ground tie and control points. In next subsection two methods 
are introduced to determine the offset of GPS, alignments and 
drift errors of IMU for high accuracy positions and attitudes of 
images. 
2.2 Generalized Bundle Adjustment 
To relate the image coordinates (x, y) of one point to its mapping 
coordinates (X,Y, Z), the following collinear equations are used: 
X XN X AX 
Y= Yn + R(ON» PN>ÆN YAR cam y + AY ) (1) 
Zl iz. =f] [AZ 
where, (Xn, Yu, Zw, On, Qu, Kw) are the exterior orientation 
parameters of the N" image line on which the image point in the 
mapping coordinate, and obtained from GSP and IMU 
observation values by removing the influence of GPS offsets, 
IMU alignments and drift errors; R(Ow, On, Kn) is rotation 
matrix of IMU to mapping coordinate system; A is scaling factor 
from image to ground; R.,, is rotation matrix of camera to 
satellite fixed coordinate system and the angles can be obtained 
from table 1 for PRISM; (AX, AY, AZ) are offsets between GPS 
antenna and perspective center of camera in satellite coordinate 
system and can be obtained from the NASDA, Japan; fis focal