\ TION
algary.ca,
Brasil
n elevation
neral. laser
(ie, object
, especially
ition of the
vever, prior
ach system
systems is
troducing a
'€ extracted
s from the
constraints
ing straight
etween the
systems. In
evealed the
ics is found
termination
n using the
and dense
's add to its
its own
information
y is worse
tacle in the
licated and
ially when
ry and laser
referable in
tage in one
1 the other.
cher quality
vever, the
only after
separately
would be
metric and
ime. (Habib
ough which
ing systems
he type and
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004
extraction of registration primitives in addition to the
registration steps required to reveal any calibration
discrepancies in the systems.
Most registration methodologies use discrete points as the sole
primitive for solving the registration problem between two
datasets. Such methodologies are not applicable to laser
scanned surfaces since they correspond to laser footprints
instead of distinct points that could be identified in imagery
(Baltsavias, 1999). Conventionally, surface-to-surface
registration and comparison have been achieved by
interpolating both datasets into a uniform grid. The comparison
is then reduced to estimating the necessary shifts by analyzing
the elevations at corresponding grid posts (Ebner and Ohlhof,
1994; Kilian et al., 1996). Several issues can arise with this
approach. First, the interpolation to a grid will introduce errors,
especially when dealing with captured surfaces over urban
areas. Moreover, minimizing the differences between the
surfaces along the z-direction is only valid when dealing with
horizontal planar surfaces (Habib and Schenk, 1999). Postolov
et al. (1999) presented another approach, which works on the
original scattered data without prior interpolation. However, the
implementation procedure involves an interpolation of one
surface at the location of conjugate points on the other surface.
Additionally, the registration is based on minimizing the
differences between the two surfaces along the z-direction.
Schenk (1999) introduced an alternative approach, where
distances between points of one surface along surface normals
to locally interpolated patches of the other surface ar
minimized. Habib and Schenk (1999) and Habib et al. (2001
implemented this methodology within a comprehensive
automatic registration procedure. Such an approach is based on
processing the photogrammetric data to produce object space
planar patches. This might not be always possible since
photogrammetric surfaces provide accurate information along
object space discontinuities while supplying almost no
information along homogeneous surfaces with uniform texture.
€
)
In this paper, the registration procedure will utilize straight line
primitives and 3D similarity transformation for aligning the
photogrammetric model relative to the laser data reference
frame. The following section previews the components of the
general registration paradigm and the particulars of applying
each component to the photogrammetric and laser datasets
under consideration. The last two sections cover the
experimental results as well as the conclusions and
recommendations for future work.
2. METHODOLOGY
The registration process aims at combining multiple datasets
acquired by different sensors in order to reach better accuracy
and enhanced inference about the environment than could be
attained through using only one sensor. The following
subsections address the components and issues necessary for an
effective registration paradigm (Brown, 1992).
2.1 Registration primitives
To register any two datasets, certain common features have to
be identified and extracted from both sets. Such features will
subsequently be used as the registration primitives relating the
datasets together. The type of chosen primitives greatly
influences subsequent registration steps. Hence, it is crucial to
first decide upon the primitives to be used for establishing the
transformation between the datasets in question. In this paper.
171
straight line features are selected for this purpose. This choice is
motivated by the fact that such primitives can be reliably,
accurately, and automatically extracted from photogrammetric
and laser datasets. The procedure adopted to extract straight
lines from the photogrammetric and laser datasets and how they
are included in the overall alignment procedure is described
below.
Photogrammetric straight line features: The representation
scheme of 3D straight lines in the object and image space is
central to the methodology for producing such features from
photogrammetric datasets. Representing object space straight
lines using two points along the line is the most convenient
representation from a photogrammetric point of view since it
yields well-defined line segments (Habib et al., 2002). On the
other hand, image space lines will be represented by a sequence
of 2-D coordinates of intermediate points along the feature.
This appealing representation can handle image space linear
features in the presence of distortions as they will cause
deviations from straightness. Furthermore, it will allow for the
inclusion of linear features in scenes captured by line camera
since perturbations in the flight trajectory would lead t
deviations from straightness in image space linear feature:
corresponding to object space straight lines (Habib et al., 2002)
Manipulating tie straight lines appearing in a group ol
overlapping images begins by identifying two points in one
(Figure la) or two images (Figure 1b) along the line under
consideration. These points are then used to define the
corresponding object space line segment. It is worth mentioning
that these points need not be identifiable or even visible in other
images. Intermediate points along the line are measured in all
overlapping images. Similar to the end points, the intermediate
points need not be conjugate, Figure 1.
1 à 1
, 3 - 4 es A a
RE e ^. Sy
: 2 i dm "M. i :
KC +=
Image 1 Image 2 Image 1 Image 2
M an 5: * s)»
X x? x n NE a
8 * 2 Ih,
bs Y o UM x À A
^ M e * NS 7 ^e 2
Image 3 Image 4 [Image 3 [mage 4
(a) (b)
e End points defining the line in object space
Intermediate points
Figure 1. End points defining the object line are either
measured in one image (a) or two images (b).
The relationship between the image coordinates of the line end
points f(x. y) (X, y and the corresponding ground
coordinates (X, Ys Z), (X.. Y Z.)j is established through
the collinearity equations. Hence, four equations are written for
cach line. The intermediate points are included into the
adjustment procedure through a mathematical constraint, which
states that the vector from the perspective centre to any
intermediate image point along the line is contained within the
plane defined by the perspective centre of that image and the
two points defining the straight line in the object space,
Figure 2. That is to say, for a given intermediate point, a
