In: Wagner W., Szekely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B
301
1. The image will be initially oriented with the
laser scanning data according to provided GCPs.
In this step the rotation angle around Z axis “k“
is only rotation angle which is considered for
further processing,
2. Equation 1 defining the relationship between the
laser scanning data and image according to
parameters of orientation will be constructed,
3. The interested features will be extracted from
image and will be transformed to laser scanning
space according to Equation 1 for matching with
their correspondences features.
4. Final transformation
For implementation of above steps correctly, a special
filter has been developed that extract the plane from laser
scanning data. The filter will remove scattered points that
interferes the matching process. For example, each laser
scanner pulse signal has three responds from trees and
bushes. These additional data affect the process of the
matching and transformation and reduce the accuracy of
outputs.
X i
3 r t
—
+ sR K
> 7 i
*t.
Z s-
Z i-
3. STUDY AREA
Test area for registering aerial image on laser scanning
data locates in Espoonlahti (approximately 60° 8’N, 24°
38’E) in southern part of Finland. The area can be
characterized as low residential urban area having mainly
terrace houses and detached houses with a multistorey
buildings and apartments in some areas.
Also the most required and applicable data were provided
by EuroSDR as well as photographs orientation
parameters and GCPs.
GCPs have been measured using real time kinematic
(RTK) GPS and some of GCPs represent on cornices of
roofs.
Laser Scanning images were acquired with Optech
ALTM 3100 and Leica ALS50-II scanners at 2005 and
2007. Technical details of imagery systems and camera
are presented below.
Optech ALTM 3100
Scanning angle 24 degrees, 20 degrees is processed
(±10°)
PRF 100 kHz
Scanning frequency 67 Hz
Flying speed 75 m/s
Leica ALS50-II
Scanning angle 40 degrees (±20°)
PRF 148 kHz
Scanning frequency 42.5 Hz
Flying speed 72 m/s
DMC Photogrammetric Images
Pixel depth 16 bit
Size 13824x7680
Ground resolution 5 cm
Forward overlap 60%
Side Overlap 20%
Interior Orientation
Focal length: 120.0000 mm (10000 pixels)
Principle point (differences from the image centre):
Px=0.000 mm
Py=0.000 mm
Pixel size: 0.012 mm
Image size: 13824x7680 pixels (165.888x92.16 mm)
The terrestrial images and 3D data have been acquired
from an ancient site in south of IRAN near city of Shiraz.
The site’s name is Naqsh-e Rostam which includes a
number of rock carvings, Graves, and a building named
Ka’ba-ye Zartosht. The images and 3D model were
acquired from Ka’ba-ye Zartosht which most of photos
were acquired from a non-metric camera.
4. IMPLEMENTATION AND EVALUATION
For implementation and evaluation of the proposed
approach, the aerial image and terrestrial images have
been split to small area according to geometry of objects.
For example all buildings on the aerial image have been
extracted and separated from the image. For terrestrial
images, all windows and the door have been extracted
and separated. Then according to Equation (1), matching
between each extracted object and 3D model or laser
scanning data has been implemented. Some objects such
as multi stories buildings have a significant depth of view
and that side wall with windows can be easily recognised.
Topography of area around Ka’ba-ye Zartosht has set a
difficulty for stereo image acquiring. The building has
been constructed by stones and has 13.5m high and only
by a convergence photography can acquire whole of each
wall. It was one of the challenging job to compile data of
each wall in a map and GIS. At first it was tried by using
a Zeiss Stereoplotter PLANICOMP P33 and a digital
photogrammetric system to compile data on a map, but
both systems failed to provide an accurate map. Then, it
was planned to obtain the 3D data by precise engineering
surveying approach. A precise 3D model has been
provided with less than 1mm STD. This 3D model has
provided precise information for analysing the output
from registering the image on the 3D model. It needs to
mention that the photos were acquired by a non-metric
film based camera. Then photos have been scanned by a
very basic scanner. The images have a very poor quality
and mostly blurred.