X-B3, 2012
mes “center of
nts are iteratively
ge extraction.
' selected before
for selected part.
he plane of quasi
: its center of
ll selected points
ited according to
it could not be
by any existing
point cloud and
of these images.
raction described
ould be used for
paring of images
most of facades
ion of geometric
tor also able to
atching fails.
n parameters of
method.
cloud geometric
mage. Geometric
atched edges are
tched vectors or
could be very
complete vector
od. That is why
erform manual
phic image. For
alculated. After
is calculated. If
ld, middle point
the same kind is
Check continues
olation of point
onto plane of
of projection to
d for each point.
operator moves
' nearest point in
rest point to the
proximate point
ed distance and
after node point
more time for
are segmented,
Depending on
ve before point
ging to one or
n the group of
Selected group
ie node point are
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B3, 2012
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia
calculated as the point of intersection between the plane and the
ray running from the center of projection through the image
plane.
3. EXPERIMENTAL RESULTS
3.1 Realisation of method
For experimental testing of the proposed method software
package was created. This package consists of two parts
implemented in C++ and in MATLAB.
Compiled C++ application allows loading of raster images
saved in device independent bitmap. If distortion parameters are
set, image transformation is performed. Transformed image
visualized on the screen for referencing and vectorization.
Point cloud loaded from PTS and PTX files — formatted text
files. After loading points are saved into special file format and
indexed. This operation is made for increasing of total
productivity of calculation and visualization. And furthermore
maximum number of loaded points in point cloud is limited
only by free hard disk space, because only needed part of data is
loaded in operative memory.
The point cloud could be visualized in 3D window. User
initiates segmentation of the point cloud and photographic
image. After that quasi image is created and passed to
MATLAB, where SIFT algorithm is implemented. SIFT gives
tie points. Tie points are returned back to application where
external orientation parameters of the photography image are
calculated using Least Squares Method. After calculating of the
orientation parameters operator is able to initiate automatic edge
extraction. Besides of that operator is able to create vectors
manually, edit vectors or delete them. Created data could be
visualized in 2D window with photography image and in 3D
window with point cloud. The result is saved in vector format.
3.2 Test site
Test object is the facade of building with regular and irregular
elements. Point cloud captured by the terrestrial laser scanner
Topcon GLS-1000. Photographic images are captured by the
digital camera CANON EOS. The camera is calibrated in
MIIGAIK laboratory.
Figure 2. Extracted edges. Quasi image (left) and photography
(right).
Figure 3. Corresponding points.
9 points out of 18 are matched, 7 of them were automatically
excluded after superposition analyses and two of them were
excluded during adjustment.
RMS of the tie points is 0.9 pixel (1 pixel is 0.0055mm).
Total number of automatically extracted edges is 800, approved
is 490. Most of them needed to be corrected.
Control measurements performed by the measuring type had
shown mean error of the created vectors below 0.016m.
4. CONCLUSIONS AND FUTURE DEVELOPMENTS
The first experimental result shown advantages and
disadvantages of the proposed method. Automatically extracted
corresponding points should be corrected manually by operator
in case of facades with mostly regular image elements. Process
of automatic tie point extraction should be improved. The
process of geometric edge extraction should also be improved
for increasing of total accuracy.
Three more test objects are prepared for future tests. It is three
different facades: simple façade with regular elements (12
images, 2.4 million points); simple with irregular elements (8
images, | million points) and complex facade of church (12
images, 10 million points).
REFERENCES
Fabris M., Achilli V., Artese G., Boatto G., Bragagnolo D.,
Concheri G., Menghello R., Menin A., Trecroci A., 2009. High
resolution data from laser scanning and digital photogrammetry
terrestrial methodologies. Test site: an architectural surface.
Proceedings of IAPRS, Vol. XXXVIII, part 3/W8, September
2009, pp.43-48
Jansa J., Studnicka N., Forkert G., Haring A., Kager H., 2004.
Terrestrial laserscanning and photogrammetry — acquisition
techniques complementing one another. The International
Archives of the Photogrammetry, Remote Sensing and Spatial
Information Sciences, Vol. 34, Part XXX
Chunmei Hu, Yanmin Wang, Wentao Yu, 2008. Mapping
digital image texture onto 3D model from LIDAR data. The
International Archives of the Photogrammetry, Remote Sensing
and Spatial Information Sciences. Vol. XXXVII. Part B5.
Beijing 2008
Ayman Zureiki, Michel Roux, 2009. Ortho-rectified façade
image by fusion of 3D laser data and optical images.
Proceedings of IAPRS, Vol. XXXVIII, part 3/W8, September
2009, pp.305-310
Nex, F., Rinaudo, F., 2009. New integration approach of
photogrammetric and LIDAR techniques for architectural