The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008
988
The statistical analysis of obtaining result showed that the
performance of implemented system is desirable to detect
building changes, while its performance have some limitations
in buildings with no changes. This is caused by detecting unreal
edge because of shadow, difference in exposure condition or
losing actual edge made by shadow or vegetation coverage.
3.1.2 Precision assessment of developed least square
matching algorithm
Precision refers to how close together a group of measurements
actually are to each other. So the accepted criteria to precision
assessment of proposed algorithm results, is shift dX. This
parameter is calculated from the equation 7. The output result
of this equation for two matched point after several iteration, is
residual shift which is considered as algorithm precision.
Indeed, the shift value is represented the ability of implemented
system to separate two edge with no change. So the shift value
of more precise system is smaller than the imprecise one. In an
ideal case of a perfect match dX=0, therefore the shift value
differences between real case and ideal one is defined the
precision of our developed system. So the precision of our
developed algorithm is calculated by averaging 10 matched
points shift values for convergence case. Experiments showed
the shift value of 0.45 pixels as algorithm precision.
4. CONCLUSIONS
The algorithm we presented in this paper extends the concept of
least squares template matching to identify object outline
changes. By using image orientation parameters and positional
data we can reduce the problem of 3-D object monitoring to a 2-
D image-space matching problem. Analysis of semantics within
a template, before the actual matching taken place, improves the
accuracy and reliability of the presented technique.
The accuracy assessment showed that the change percentage of
the regions that the algorithm can detect them correctly is 70%,
the change percentage of the regions that the algorithm can not
detect their changes is 10% and the change percentage of the
regions with no changes while our algorithm detects changes in
these regions is 20%. The statistical analysis of obtained result
showed that the performance of implemented system is
desirable to detect building changes, while its performance has
some limitations in buildings with no changes. This is caused by
detecting unreal edges because of shadow, difference in
exposure condition or losing actual edges made by shadow or
vegetation coverage.
The ability of implemented system to separate two edges with
no change is considered as precision criteria. Indeed, the
average of shift values represents the precision of our
implemented system. Experiments showed that the algorithm
precision is 0.45 pixels.
The capability of this novel approach is representing the image
and object coordinates and change percentages of urban area
and also each building in it.
Bohuslav,V.,2004, "Image matching and its applications in
photogrammetry", 15-53.
Gruen A., 1985,"Adaptive Least Squares Correlation: A
Powerful Image Matching Technique", South African Journal of
Photogrammetry, Remote Sensing & Cartography, Vol. 14, No.
3, pp. 175-187.
Agouris P.,1992, "multiple image multiple matching for
aerotriangulation". The Ohio state university.Colubbus.ohio.
Baltsavias , E.P. , 1991,",multiphoto geometrically constrained
matching" 49. ETEl-institute for geodesy and
photogrammetry.Zurich.
Habib, A., and Kelley, D. ,2001,"Automatic relative
orientation of large scalelmagery over urban areas using
Modified Integrated Hough Transform", ISPRS Journal of
Photogrammetry& Remote Sensing 56, pp. 29-41.
th
Kraus, K., 1997, "Photogrammetry", Vol.2, 4 edition, Fer.
Diimmlers Verlag, Bonn, ISBN 3-427-78694-3.
Grien, A., and Akca, D., 2005, "least square 3D surface
matching", ASPRS annual conference, March 7-11.
Phalke, S., May 2005, "Change Detection of Man
made Objects Using Very High Resolution Images",
university of CALGARY, Department of Geomatics
Engineering.
REFERENCES
Geodaetischas seminar ss/2000. "Matching methods for
automatic DTM generation".
www.photogrammetry.ethz. ch/general/person/maria/matching.p
df((accessed December 7,2007).