International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
1.3 Available Data
The wider area of the National Technical University of Athens
(NTUA) campus was covered by 1:3500, 1:6000 and 1:8000
image scales taken on the same day (08/08/1999). The B&W
images present forward and side overlaps of 60% and 25%
respectively. It is a slightly hilly area with a big mountain as its
east boundary. The camera used was a Z/I RMK TOP 15 with
152.61 1mm principal distance. The analogue negatives were
scanned at a 1800dpi resolution, i.e. pixel size of 14um using a
photogrammetric scanner. This high resolution would ensure
good and straightforward recognition of the control points
(Ackermann & Krzystek 1997). The Z/I ImageStation digital
photogrammetric workstàtion was employed for this project.
The mensuration software offered are the ISDM (Image Station
Digital Mensuration) and the ISAT (ImageStation Auto
Triangulation) for the interactive and automatic triangulation
respectively. The number of images per strip and scale are
shown on Table ! and Figure 2 presents an aerial image of the
area, in which the campus of the National Technical University
of Athens appears.
Image scale | Images/strip | Strips Total of
Images
1:3500 7 2 14
1:6000 5 2 10
1:8000 4 ] 4
Table 1: Image data
ia ded qe
Figure 2: Image scale 1:3500, with N.T.U.A. depicted on it
A number of control points were measured using a pair of GPS
receivers to an accuracy of a few centimetres, in the Greek
Geodetic Datum (EGSA87) a Transverse Mercator projection
with GRS80 as reference ellipsoid.
1.4 Image Station Automatic Triangulation (ISAT)
Image Station Automatic Triangulation (ISAT) is an automatic
aerial triangulation software package equipped with powerful
editing facilities. The objective of ISAT is to automate point
transfer and tie-point measurement operations, thereby,
minimizing manual work and operator intervention and
596
maximizing objectivity and speed. In addition, ISAT supplies
editing facilities with Z/I Imaging’s Image Station
photogrammetric products.
1.5 General Workflow of ISAT
The approach is intended as a fully automatic process. The
entire procedure is characterized by two main steps:
1. Initialization, which determines enough, fairly accurate
tie point areas at the von Gruber point positions.
2. Application of the matching strategy in the homologous
image patches through the image pyramid by the kernel
system.
Preparation
e parameter editing :
e Interior orientation
e measurement of control points
Initialization of tie point areas
With consideration of DTM, GPS,INS
om 9 m
i Visual check / Edit |
|
| Derivation of pyramids
Kernel System
e Feature extraction
e Preliminary matching
e Automatic point transfer in robust
bundle adjustment
e DEM refinement
(Feature Based and
Least-Square Matching)
€
| Visual check / Edit | S
| Final block adjustment | &
9 Auto pe
Figure 3: General workflow of ISAT
Interactive
The ISAT process (Figure 3) applies image pyramids for a
coarse-to-fine strategy. In each pyramid level the following
steps are performed:
e Automatic initialization of the tie point areas. Tie points are
preferably selected at the vonGruber positions, where the
best multi-image tie points are located.
e Feature extraction using Fórstner operator (Fórstner 1986)
e Preliminary image matching using a combination of feature
based and least squares matching and a sophisticated
strategy for obtaining evenly distributed tie points over the
whole block.
e Robust bundle block adjustment for automatic tie point
selection and for the initialization of the next pyramid level.
e [n the last pyramid level the internal block adjustment
supplies the final adjustment result.
In weak block areas tie points may be added by interactive
point measurement. After manually adding points, the internal
Intei
———
bloc
auto
Hav
over
form
thro
mat
likel
corr
matc
refer
imag
valu
the 2
equa
matc
suce
The
autor
1994
matc
(Jaak
In ac
and |
the r
and t
2.1
For |
with
solut
imag
1:351
Was
1:60
1:80(
The 1
as co
aver:
deviz
orien
the o
