International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004
further‘ processing. The hierarchical data structure is used for
both the range measurement and camera data. A series of co-
ordinate systems which define the geometric dependence
between the sensor's raw data to a higher order reference
system (for details see Ullrich et al., 2003) is the basis of all co-
ordinate transformations. The following list provides an
overview of the various local co-ordinate systems:
— SOCS (Scanners own co-ordinate system) is the
coordinate system of the raw data where the polar
measurements are based on. It is defined by the
rotation axis of the scanner unit (=the origin for the
angular measurements) and the reference direction
(i.e. internal 0-azimuth)
— PRCS (Project co-ordinate system) is the central
reference system for a Riegl laser scanning project. It
is as local system as far as the project area is
concerned whose co-ordinate range should not exceed
10 km.
— GLCS (Global co-ordinate system) is hierarchically
above the PRCS. It is usually the co-ordinate system
of a national reference system.
_ CMCS (Camera co-ordinate system) is the reference
of the camera mounted on top of the laser device.
Mounted
Rotating
Scanner
Head
Scanner Base
Fig.l: Scanner and Camera Coordinate System
Co-ordinate Systems
i
i
i
iS =
1
|
[^
|
| I
| Scanner | Camera
| Geb | Project -
L orem
T Const.per | Const. per = Const. per. |
Project Device Position Scan Azimuth o
Mor \ aM. |
Const. per Const. per
“Scan Azimuth [Camera Mount.
ASA
Moor = No) Mount |
Fig.2: Co-ordinate Systems and Transformation Matrices
The two transformations with their respective matrices Mgop
and Mpop describe the transition from raw laser data to the
global co-ordinate system via the project co-ordinate system. In
the case of camera data, an additional transformation has to be
applied, i.e. the transformation of the camera system into the
scanner system. One should keep in mind that the scanner body,
and therefore also the camera mount, is rotating during the
range measurement process and thus the camera system also
rotates with respect to the SOCS (Fig.1). This transformation
has been realised by splitting up this time-variant step into
(Fig.2):
_ The time-variant rotation of the range measurement
unit (i.e. the body) with respect to the 0-azimuth
reference direction, expressed by the matrix Mcop,
and
— By the time-invariant rotation and translation of the
camera system with respect to the SOCS, expressed by
the so-called mounting calibration Mu
The following two equations are used for the transformation of
the measured co-ordinates from the scanning system to the
global system and from the camera system to the global system,
respectively:
Xaues = Mror Msop.n XSOCS,n
a A T! :
XGLCS — Mpor Msop.n Mcop.nm M mount XCMCS.n,m
where n denotes a certain position of the scanning device in the
project area and m denotes (at the device position n) a certain
camera position (determined by the azimuth angle Q0). when the
photo has been taken. Note that Mu is constant as long as the
camera has not been moved with respect to the scanner body,
i.e. not been detached from the scanner body and mounted
again. The great advantage of the stable mounting calibration is
the possibility to reference camera images to the laser scanner
data at any stage of the data acquisition as the geometric
relation between each laser point and the camera direction is
known at any time. (Note that the azimuth angle is delivered
by the laser unit for each measured laser point.)
2.2 Orientation using Laser Scans and Images
The determination of the Mgop matrices is the central task of a
laser scanner project where usually an arrangement of many
scanning positions is necessary to cover the whole area of
interest. The individual scanning positions are tied to each other
with the help of tie points through a block adjustment. RiScan
Pro provides appropriate means to built a homogeneous block
and determine the transformation matrices. Tie points can be
identified in the intensity image of the range measurements (Le.
the intensity of the received echo pulse), they may be measured
in the camera images, or in both. A combination of both
techniques is to be preferred, as it bears the potential to reduce
the time for measurement and to improve the accuracy. On the
one hand, laser measurement can support measurement of
targeted tie points in the images, on the other hand both sets of
measurement may be input to the block adjustment procedure in
order to determine all unknown orientation parameters of the
set-up and, if necessary, even the instrument calibration in one
step of a hybrid bundle adjustment.
2.3 Reconstruction of the Object
As for the object reconstruction the great advantage of camera
images is their relatively high spatial and spectral resolution. In
particular coloured object features, small details and prominent
object discontinuities, such as edges and corners, can be
identified more accurately in photographs than in range images.
The disadvantage of reconstruction from photographs is the
high effort and quite often low reliability if automatic pro-
cedures are employed, especially in close-range applications.
Image matching algorithms usually need good approximations
before accurate fine measurement can commence. The range
measurements of the laser scanner deliver 3D object points and
thus very good first approximate positions in the images. The
automated reconstruction process works much faster and is able
to provide more reliable, more complete and possibly even
more accurate results.
One metric product which can be created very quickly and
without great effort is a Z-coded True Orthophoto (ZOP) of an
object area. This sort of orthophoto is especially useful if
950
In
bi
fo
th
or
th
th
se
of
pl:
a
lik
mM
Th
tox
cal
vi:
tru
de
sci
pr
str
are
mc
im
a v
hay
Th
vis
For
clot
hug
aut
bee
Seve
Scar
The
met
