![]()
The ITG digital terrain model generation systemITG (O'Neill and Denos, 1996, Denos 1991, Otto and Chau, 1989) is a fully automated system which generates digital terrain models [DTM's] fromoptical stereo imagery. Currently, the system works with SPOT imagery (O'Neill, 1992). A Tsai (Tsai, 1987) central perspective camera model has also recently been added to the system. This allows it to be used with aerial photography. The ITG system has been used by Geophysicists in Oxford Department of Earth Sciences to characterise terrain in the area of active fault zones in Greece and Turkey. The system has alsobeen used by the Cardiff Geophysics Consultancy terradat to in both gravity surveys and intervisibility studies for windfarms.
Components of the ITG terrain model generation system
The autoseeder is used to generate a set of (approximate)
stereo correspondences
which are used to seed the optical stereo matcher. The autoseeder is divided
into two parts; cascade which uses an
annealer to generate a small number of random stereo correspondences and
cheops which applies a resolution pyramid
to extract those stereo-correpondences which correpond to actual matches.
These stereo correpondences are used to seed the pslam
area correlation stereo matcher. This is sheet growing adaptive least squares pattern correlator
which is based on the work of Prof. A. Gruen (ETH Zurich) and Paul Otto and Tony Chau (Alvey
MMI-137 Project, University College London). Pslam extends the basic sheet growing ALSC
[Adaptive Least Squares Correlation]
technique, enabling it to work tolerably well on images which are:
The stereo matcher produces a dense array of stereo-correpondences. These are either output producing
a digital disparity model, which can be further processed using tools like
disp2ddm or they are passed to the
camera model which transforms them from image
space to ground space, producing an ungridded array of (x,y,z) co-ordinates in a geocentric
co-ordinate system. These points are converted to an appropriate local vertical system using the
co-ordinate conversion filters. Currenlty filters are
available to transform between geocentrics and:
Finally a regularly gridded
digital terrain model DTM is produced by
interpolating
the set of points produced by the co-ordinate conversion filters. The interpolator used by the
ITG system, krige_spb, is based on the
Kriging interpolator
(Delfiner and Delhomme, 1975) implemented by
Tim Day for the Alvey MMI-137 Project at
University College London (Day 1990). It has been extensively modified to both improve
it throughput and to allow it to efficiently interpolate datasets containing millions of points on
mid range
Pentium
Personal computer's and similar hardware. In addition to the Kriging interpolator a simple
nearest DTM cell gridding algorithm fndem is
also available. This is typically used to generate quick look DTM's at coarse resolution.
In addition to producing DTM's, a pipeline generator gti
is provided which generates
orthoimages using the DTM generated by the ITG subsystem and
one of the images (used by ITG to generate the DTM. gti automatically sets up an orthoimage
generation pipeline consisting of the
fnxyz,
spot1m (inverse),
cocon (inverse),
npraw filters.
ITG may also be used in stand alone mode a change predictor. Coupled with appropriate
imagery from a scanning radiometer, for example the ATSR-1 instrument carried by the
European Remote Sensing Satellite,
ERS-1 or
the Japanese
ASTER satellite (O'Neill and Dowman, 1993),
it can function as part of an automated system to assess the health of large-scale ecosystems.
![]()
Obtaining the ITG stereo matching software
If you are interested in more information about ITG contact
Mark O'Neill
(mao@tumblingdice.co.uk),
ReferencesDay 1990: Day T, Semi-variograms Kriging and Stereo Matching Alvey MMI-137 Final Report, Chapter. 18. Delfiner and Delhomme, 1975: Delfiner P. and Delhomme J.P, Optimum Interpolation by Kriging, in Display and Analysis of Spatial Data J.C Davis and M.J. McCullagh (Eds), pp.96-114, Wiley. O'Neill 1992: O'Neill Mark A, A Kinematic Model of the SPOT-1 Sensor, Ph.D Thesis, University of London. O'Neill and Denos 1996: O'Neill Mark .A and Denos Mia, Automated System for Coarse-to-Fine Area Corellation Stereomatching Image & Vision Computing 14(3):225-236. Denos 1991: Denos Mia, A pyramidal scheme for stereomatching SIR-B imagery Int. J. Rem. Sens. 13(2):387-392. O'Neill and Dowman 1993: O'Neill, Mark A. and I.J. Dowman, A simulation study of the ASTER sensor using a versatile general purpose rigid sensor modelling system (Advanced Spaceborne Thermal Emission and Reflection Radiometer). Int. J. Rem. Sens., 14(3): 565-582. Otto and Chau 1989: Otto G.P. and Chau T.K.W, Region-growing algorithm for matching of terrain images Image & Vision Computing 7(2):83-94. Tsai 1987: Tsai R. Y, A versatile Camera Calibration Technique for High-Accuracy 3D Machine Vision Metrology Using Off-the-Shelf TV Cameras and Lenses, IEEE Journal of Robotics and Automation 3(4): 323-344 AcknowlegmentsMany people have contributed to testing the ITG system in its pre-commercial phase. The principal contributors are:
Barry Parsons (Earth Sciences, Oxford University, UK)
Content (c) 2007 Tumbling Dice Ltd. ITG is a Tumbling Dice Ltd product. |