International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
contrast boundaries between Lawns and Forest areas
resulted in these regions being merged into a single
patch. The confusion between Building and Road was
not a result of segmentation as generally these two
classes were well delineated. However, confusion
occurred because the spectral radiances of the two
classes were sometimes very similar. This arises
because materials such as asphalt, stone and concrete
are used for both building roofs and roads.
As part of the classifications carried out using
maximum likelihood, all pixels were assigned to the
class with the highest likelihood. This is a relative,
not an absolute measure. Thus even classes that result
in very low likelihood when compared to all the
training data sets are classified. It is possible that a
region is not represented by any of the training data
sets, and this should be identified. In future work, it
may be desirable to establish an absolute minimum
maximum likelihood for classification. Patches that
fail to meet the minimum value would be flagged as
unknown.
6. CONCLUIONS
This study produced a region-based classification
approach specifically designed for high spatial
resolution imagery. The new classification method
resulted in improved results at both the image object
scale and a richer attribution at the aggregate land cover
scale. This research made a contribution to the growing
field of analysis of high spatial resolution imagery.
The methods developed in this research are important
not just because they produce more accurate results that
show the spatial patterns more clearly because of their
lack of distracting high frequency noise. The
delineation and attribution of image objects, rather than
classified pixels, is an important step toward integrating
remote sensing with GIS. The object-based approach
resulted in a pleasing simplicity of spatial structure
compared to the noisy patterns of traditional pixel-
based classification.
Acknowledgements
This Project was funded by the Ministry of Science and
Technology, Republic of Korea.
References
Cao, C., and N. S. Lam, 1997. Understanding the scale
and resolution effects in remote sensing and GIS. In:
Scale in Remote Sensing and GIS (D. A. Quattrochi,
and M. F. Goodchild, editors), Lewis Publishers, New
York, NY, pp. 57-72.
Eastman, J., R., 2003. IDRISI Kilimanjaro Guide to
GIS and Image Processsing, Clark University,
Worcester, MA.
ERDAS, 1999. ERDAS Field Guide, ERDAS,
Atlanta, Georgia, 672 p.
Gougeon, F. A. 1995. Comparison of possible
multispectral classification schemes for tree crowns
individually delineated on high spatial resolution MEIS
images. Canadian Journal of Remote Sensing 21(1): 1-
9.
Janssen, L. L. F., and M. Molenaar, 1995. Terrain
objects, their dynamics and their mornitoring by the
integration of GIS and Remote Sensing. [EEE
Transactions on Geoscience and Remote Sensing 33
(3): 749-758.
Jensen, J. R., 1996. Introductory Digital Image
Processing, Prince Hall, Upper Saddle River, New
Jersey. York, NY, pp. 3-78.
Rettig, R.. L., and ,D- A. Landorehe, 1976.
Classsification of multispectral image data by
extraction and classification of homogeneous objects.
IEEE Transactions on Geoscience Electronics GE-
14(1): 19-26.
Latty, R. S., R. Nelson, B. Markham, D. Williams, D.
Toll, and J. Irons, 1985. Performance comparisons
between information extraction techniques using
variable spatial resolution data. Photogrammetric
Engineering and Remote Sensing 51 (9): 1495-1470.
Lillesand, T. M., and R. W. Kieffer, 1994. Remote
sensing and image interpretation, John Wiley & Sons,
Inc. New York, NY, pp. 169-178.
Marceau, D. J., D. J. Gratton, R. A. Fournier, and J.
Fortin, 1994a. Remote sensing and the measurement of
geographical entities in a forested environment. 1. The
scale and spatial aggregation problem. Remote Sensing
of Environment 49: 93-104.
Marceau, D. J., D. J. Gratton, R. A. Fournier, and J.
Fortin, 1994b. Remote sensing and the measurement of
geographical entities in a forested environment. 2. The
optimal spatial resolution. Remote Sensing of
Environment 49: 105-117.
Mcdevitt, R. J., and S. D. Peddada, 1998. An automated
algorithm for cleaning and ordering the boundary
points of a one-dimensional curve in a segmented
image. /EEE Transactions on Geoscience and Remote
Sensing 36 (1): 307-312.
Meyer, P., K. Staenz, and K. I. Itten, 1996. Semi-
automated procedures for tree species identification in
high spatial resolution data from digitized colour
infrared-aerial photography. /SPRS Journal of
Photogrammetry & Remote Sensing 51: 5-16.
Pax-Lenney, M., and C. E. Woodcock, 1997. The effect
of spatial resolution on the ability to monitor the status
of agricultural lands. Remote sensing of Environment
61: 210-220.
Price, J. C., 1994, How unique are spectral signatures?
Remote Sensing of Environment 49: 181-186.
Swain, P. and S. Davis, 1978. Remote Sensing: The
Quantitative Approach, McGraw Hill, N.Y. 369 p.
Teillet, P. M., K. Staenz, and D. J. Williams, 1997.
Effects of spectral, spatial, and radiometric
characteristics on remote sensing vegetation indices of
forested regions. Remote Sensing of Environment 61:
139-149.
Tou, J. T. aud R. C. Gonzalez, 1974 "Pattern
Recognition Principles, Addison-Wesley Publishing
Co, Reading, MA.
KE
AB
Air
ger
poi
obj
she
nev
me
for
cor
me
for
Air
acq
For
dat
uni
red
Sur
iter
mo
19%
The
det
mo
alg
filte
2.1
Dig
dis
arti
dig
tha
exa
Ho
sen
loc
rep
dis
poi