rs to
d for.
efits,
ound
tions
exhi-
rid
Nide-
VWW
'orm-
s and
S the
ucts.
'Okes
new
ions.
elop-
art of
n the
some
used
lica-
5, the
iccu-
Norld
Sys-
liffer-
and
new
es in
com-
osent
tured
date
hard-
rious
rated
f giv-
envi-
n 2000.
TU12 - Multi-source data fusion in photogrammetry
and remote sensing
Convenor: C. Pohl
Location: ITC Enschede
Date: July 14th, whole day
Lecturers: C. Pohl,
E. Baltasvias,
M. Hahn
Number of participants: 16
This tutorial provided an introduction to multi-source data
fusion. It dealt with fusion of sensor data and other knowl-
edge sources (databases, model knowledge, rules etc.),
fusion of multi-sensor data at different processing levels. It
covered both airborne and spaceborne sensor data. After
introducing definitions, an overview of multi-source data
fusion was given. This included possible applications, data
types involved, different fusion levels and processing tech-
niques, problems and requirements. The tutorial then con-
tinued in two major blocks:
Fusion of image data and GIS/map databases and fusion
of multiple cues, including topics like fusion/homogenisa-
tion of GlS/map databases.
Data, especially image fusion, using mainly data from
spaceborne sensors.
Both blocks provided an overview on data, fusion tech-
niques, processing chains and knowledge requirements,
followed by illustrative examples. Special emphasis was
placed on examples and case studies using satellite data
(SPOT, Landsat, ERS-1/2, JERS-1, IRS-1C) and airborne
sensors (colour and NIR film cameras, stereo and multi-
spectral line CCDs, laser scanner). Examples of extraction,
reconstruction and classification of objects were pre-
sented, with applications in topographic mapping and
map updating, change detection, generation of 3-D city
models, land use monitoring and natural disaster manage-
ment. The presentation included selected elements out of
a multimedia tutorial on data fusion developed by ITC.
ISPRS
2000
TU15 - Image analysis techniques for aerial image
interpretation
Convenor: W. Eckstein
Location: RAI Amsterdam
Date: July 15th, afternoon
Lectureres: W. Eckstein,
C. Steger
Number of participants: 19
First, the advantage of multiple input data (e.g., colour,
infrared, DEM) and the information that can be derived
from these sources was discussed. Besides sensor data,
‘synthetic” input images were generated (e.g., using tex-
ture filters) to support the segmentation process.
Depending on the class of the object, the extraction must
be selected: compact artificial objects may be seg-
mented using primitives like areas, lines or points. Linear
objects, like roads, are similar but the borders are curves
and their size is not limited. Arbitrary areas like meadows,
forests or fields have an arbitrary boster and are mainly
defined by their specific texture. Objects like trees or cars
have to be treated in a very specific manner. Finally, dif-
ferent base algorithms for segmentation were discussed:
pixel classification is very simple but lacks the use of
context. The extraction of primitives (edges, lines, area,
points) may be used as a basis for a wide class of
objects. Texture analysis can be used for a rough seg-
mentation of the image. Specialised operations are use-
ful for the extraction of objects like single trees or to sup-
port the interpretation process.
In the second part, different examples of image interpre-
tation were given. These comprised road, forest, and
building extraction. Here the use of features, knowledge,
models and control structures was discussed.
Tutorial
TLI1
TU2
TUS
TUS
1012
TU15
Total
Participants
Duration (days) |
18 0,5
41 1,0
12 1,0 |
9 0,5 /
16 1,0 /
19 0,5
115 /
International Archives of Photogrammerty and Remote Sensing. Vol. XXXIII, Part A. Amsterdam 2000.
129
