sed laser
reflected
ulses are
je veget-
er to the
e terrain
ABSO-
r scanner
ocess the
of meas-
absolute
sensor at
e ground
y the ab-
of meas-
(INS) or
ral GPS
in be ob-
le sensor
mination
ristics of
ning and
attitudes,
) provide
nt. The
) 10 Hz),
(1000 Hz
f the air-
meters is
sitioning
rally rel-
d with a
R
nation of
distance
D meas-
available.
red data
Roughly
nguished
inates of
vided by
'efore the
and laser
and atti-
measure-
be trans-
formed into the ground coordinate system.Using the ori-
entation parameters (Xo, Yo, Zo, w,«, &) and the measured
range Sr, the 3D coordinates (Xr,YrL,Zr) of a specific
laser footprint are computed by
XL Xo 0
Yn =| % | +Rlv,p,r) 0
ZL Zo Sr
3.2 Calibration of the laser sensor system
To cover larger areas, several overlapping strips have to be
measured similar to an aerial image flight. Due to uncor-
rected systematic errors of the single sensors, especially of
the GPS and INS sensors used to provide position and ori-
entation of the laser scanner on the installation angles of the
laser sensor, the single overlapping strips do not fit to each
other exactly. Therefore a calibration of the laser sensor
system is necessary. The aim of the calibration procedure
is to determine additional transformation parameters (cal-
ibration parameters) for the transformation of the single
strips to a homogeneous exterior coordinate system.
The overlapping areas of the single strips can be used to
perform the calibration task. Therefore a two step method
is applied: in a first step tie- and control point information
is determined and in a second step this information is used
to connect the single strips to each other and transform the
block to an exterior coordinate system. Figure 2 shows the
principle of the procedure.
qe RS
Eu P em "o —m
ug n uw NL S |
WW Wo Wer
=
strip 2
w
s em
>
>
| | control-DEM E tie-DEM —> estimated translations
Figure 2: Tie and control points for DEM matching
Determination of tie- and control point information
Due to uncorrected systematic errors two neighboring laser
strips do not fit together exactly or the single strips do not
fit to the exterior coordinate system, i.e. the single strips
do not refer to the same (exterior) coordinate system. The
aim of the calibration procedure is to determine transform-
ation parameters for each strip to allow the transformation
of each measured laser point to an exterior coordinate sys-
tem.
In the first step of the calibration procedure tie- and con-
trol point information is determined. Therefore a match-
ing process is applied to estimate translation parameters
between corresponding areas of two different digital el-
evation models. The translation parameters between two
identical windows of different laser strips are used as tie
point information, the translations between windows of
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
laser strips with external information, covering the cor-
responding area, are used as control point information.
To provide external information for the position control,
e.g. ground plans of buildings can be used by measuring
corners of buildings in the ground plans and the corres-
ponding corner coordinates in the laser data. Points in flat
areas, e.g. street crossings, can be used for height control.
The matching of laser data with external digital elevation
models is another possibility to provide control point in-
formation.
For each pair of windows in the overlapping areas three
translation parameters d.X , dY , dZ are determined. There-
fore an algorithm, originally developed for intensity based
image matching [Haralick & Shapiro 1993] was adapted for
matching height data. The original algorithm, using in-
tensity images (grey values in matrix form) was modified
to handle irregular distributed height data. The matching
process is a three step procedure, determining approximate
values of the unknowns in a first step, translations in .X
and Y between two corresponding windows in the second
step and the height offset (Z translation) in the last step.
In the second step occlusions resulting from the different
view points of the scanner are eliminated to get a better
matching result. The height offsets are exclusively estim-
ated in flat areas. Even though a affine transformation (7
unknowns) is assumed for the matching process, only the
3 translation parameters were used for the second step of
the calibration process.
Figure 3 shows the results of a matching process for the
overlapping area of two different strips. The data set was
measured with a prototype version of the sensor described
in table 1. This prototype uses 64 pixels per scan. The
flight height was 300 m and the measurement frequency
was 300 Hz. This resulted in a point distance of about
30 cm in flight direction and about 3 m perpendicular
the flight direction. For this test the positioning and atti-
tude determination of the laser rangefinder was only per-
formed with an Inertial Navigation System, which resulted
in rather large offsets and drifts in the estimated transla-
tion parameters between two different strips as shown in
figure 3. Each point in this figure, consisting of dX, dY,
dZ , can be considered as a tie point information.
6 |
[m] |
à &
0
-2-
4
-6
-8 me
: mDX DY eDZ Simic]
210 um 1 L —À
16 170 475 180 185 190
Distance between two windows: 10m
windowsize: ~40m * 40m
Figure 3: Matching results before calibration
Estimation of transformation parameters
Using the tie and control point information a strip adjust-
ment is performed. The result of this step is a set of trans-
formation parameters for each single strip. This transform-
ation parameters allow a transformation of each measured