ISPRS
2000
Terms of Reference:
Single- and multi-sensor system calibration;
Automatic sensor orientation using GPS and INS infor-
mation, and digital image processing techniques;
Automatic aerial triangulation;
Combination of sensor orientation and calibration algo-
rithms with 3D scene reconstruction algorithms;
Multi-sensor information integration, e.g. aerial and
satellite imagery, radar data, multi-spectral imagery, in
conjunction with WG 11/2
WG IIl/2 Algorithms for Surface Reconstruction
Chair Amnon Krupnik
Co-chair Charles Toth
Secretary Maxim Fradkin
Terms of Reference:
Algorithms for surface reconstruction from different
cues, such as stereo, shading, shadows, texture, and
spectral properties
Integration of information from different sources,
including aerial and satellite imagery, radar data, range
data, multiple images and GIS data
Automatic verification and quality control of recon-
structed surfaces
Interrelations between matching and interpolation
Role of surface reconstruction in automatic scene
analysis
WG IIl/3 Feature Extraction and Grouping
Chair Helmut Mayer
Co-chair Ram Nevatia
Secretary Albert Baumgartner
Terms of Reference:
Segmentation of images into meaningful features, such
as points, edges and regions
Investigation of the impact of geometric and radiometric
image formation on image segmentation
Aggregation of features by iterated grouping methods
Verification of global consistencies by optimisation
techniques
Investigation of the potential of performing grouping
processes in 3D under special consideration of object
space constraints, precision and reliability
Evaluation of the performance of feature extraction and
grouping algorithms in co-operation with similar efforts
in computer vision
WG III/4 Image Understanding/Object Recognition
Chair Wolfgang Eckstein
Co-chair Eberhard Gülch
Secretary Carsten Steger
Terms of Reference:
Theoretical/conceptual investigations in 3D object
space recognition and image understanding
Spatial modelling of man-made and natural objects like
buildings, roads, or vegetation (geometric and seman-
tic modelling)
Theory and algorithms of automation of 3D object
recognition and object extraction
Automatic model generation from CAD data and
example images
International Archives of Photogrammerty and Remote Sensing. Vol. XXXIII, Part A. Amsterdam 2000.
- Generic control structures
- Knowledge engineering
- Integration of multiple sources (DEM, GIS, colour, etc.)
in co-ordination with WGIII/5
- Performance characterisation of algorithms (quality
control, test procedures), jointly with WGIII/3
WG IIl//5 Remote Sensing and Vision Theories for
Automatic Scene Interpretation
Chair Beata Csatho
Co-chair DeLiang Wang
Secretary Erzsebet Merenyi
Terms of Reference:
- Merging theories from remote sensing (classification)
and computer vision for Object Recognition/ Image
Understanding, to analyse and interpret multi-spectral,
multi-sensor imagery
- Exploration of the potential of scale space theory on
multi-spectral imagery; extending the theory by includ-
ing the spectrum
- Combination of modelling efforts of real world objects
in GIS, digital photogrammetry and computer vision
- Investigation of the possibility of constraining the 3D
image understanding and object recognition problem
by GIS information, in co-operation with IC WG IV/III.2
and WGIII.4
WG Ill/6 Theory and Algorithms for SAR
Chair Laurent Polidori
Co-chair Soren Madsen
Terms of Reference:
- Theory and algorithms for SAR, particularly for interfer-
ometry and feature extraction
- Establishment of state-of-the-art radar mapping
- Geometric relationships in processing and interpreting
SAR data
Quality control aspects of radar images and derived
products, such as radar maps and DTMs
3. Workshops Organised under the Auspices of
Commission III
Commission Ill addresses a wide range of activities, from
data acquisition, surface reconstruction, object recogni-
tion to image understanding. The State of Science in these
diverse topics can be judged from papers presented at the
following workshops, organised by the working groups of
Commission Ill.
The most important trends that emerged during the 4 year
period may be summarised as follows:
- Sensor orientation enjoys new research interest. Two
major trends can be observed. First, the mathematical
model is extended to allow for direct solutions, increas-
ing the flexibility of data acquisition and reducing the
need for good approximations. Second, there is a trend
towards including features as entities into the orientation
process. This is a significant step forward because fea-
tures are more robust entities than points.
- The workshop in Barcelona, Direct vs. Indirect Orienta-
tion (see Table), provided an excellent snapshot of the
status of direct platform orientation systems that have
reached a high level of performance. But it also
showed limitations. While some of these limitations will
2187