International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004
Figure 6. Upper: Left- Landsat 7 Pan (15m) and MS (30m)
43,2 band merge image (310 x 315 pixels); Right-
classification result. Lower: Left- SPOT 5 Pan (5m) and XS
(10m) 1,2,3 band merge image (929 x 943 pixels); Right-
classification result.
Figure 6 shows the results of unsupervised classification of
Landsat 7 ETM+ and SPOT 5 HRG merged images. Landsat
7 merged images are clustered into 5 clusters with ISODATA
algorithms, then classified into three landscape/landcover
classes. In case of SPOT 5 merged image, 9 clusters were
generated using ISODATA algorithms, then categorise into
three landscape/landcover classes. SPOT 5 merged image
classification results in more detail information and more
sharper object boundaries than Landsat 7 merged image
classification. However, SPOT 5 merged image produces
more spatial heterogeneity. Visual inspection and existing
maps are used to examine the classification results.
Result of Landsat 7 merge image
classification (Pan and band 4,3,2 combination); Right:
Result of SPOT 5 merge image classification (Pan and band
12:3).
Figure .7.: Left:
Figure 7 shows the results of supervised classification of
Landsat 7 and SPOT 5 merged images for linear feature and
water body feature extraction. SPOT 5 merged image provide
better accuracy in extracting channels and open water bodies.
Very sharp boundaries of these features can be extracted.
Due to low spatial resolution, Landsat 7 merged image don't
300
provide detail spatial information in comparison to SPOT 5
merged image. The accuracy of classification is evaluated
using visual inspection, field knowledge as well as existing
map information. "
S. CONCLUSIONS
Satellite remote sensing images provide a valuable tool for
identification and characterisation of wetland features and
related land cover types. The synergistic use of images from
different sensors with varied spatial and spectral resolutions
have the potential for better extraction and classification of
features. Extraction and classification of landscape and land
cover features from multispectral images as well as different
resolution merged images of Landsat 7 ETM+ and SPOT 5
HRG sensors over the Okavango delta shows that
classification accuracy and detail information content
increases with increasing spatial resolution. Thematic
information extraction was carried out using supervised and
unsupervised classification to produce wetland landscape/
land cover classes for different spatial resolution data set.
The results indicate that as spatial resolution increases, high
spatial frequency landscape/land cover features are extracted
in increasing detail. However, increase in spatial resolution
also increases spatial heterogeneity. Further investigation
need to be carried out by extracting texture features from
panchromatic images and merging them with multispectral
images. The segmentation and classification of texture
merged image have potentiality to increase classification
accuracy.
ACKNOWLEDGEMENTS
The author thanks the Department of Water Affairs of
Botswana for highly effective support during field campaigns
and ETH Zurich for financial support.
REFERENCES
Anderson, J.R., Hardy, E.E., Roach, J.T. and Witmer, RE,
1976. A land use classification system for use with remote
sensor data. U.S. Department of the Interior, U.S. Geological
Survey Professional Paper 964, Washington D.C.
Ashton, P. and Neal, M., 2003. An Overview of key
strategic issues in the Okavango basin. In: Tranboundary
River, Sovereignty and Development: Hydropolitical Drivers
in the Okavango River Basin, Turton, A., Ashton, P. and
Cloete, E. (Eds.), AWIRU & Green Cross International,
Pretoria & Geneva, pp. 31-63.
Bailey, A., 1998. Okavango: Africa's Wetland Wilderness.
Struik Publishers (Pty) Ltd., Cape Town, SA.
Forstner, W., 1992. Fusing information in remote sensing.
Invited paper, XVII ISPRS Congress, Washington D.C.
Hughes, F.M.R., 1996. Wetlands. In: The Physical
Geography of Africa, Adams, W.M., Goudie, A.S. and Orme,
A.R. (Eds.), Oxford University Press, Oxford, UK, pp. 267-
286.
International
TC CL.
McCarthy, T
W.N. 1998.
Botswana —
African Jourr
Pohl, C., 199
different spa
Photogramme
Vol. 32, Part
Pohl, C. and
fusion in
applications.
pp. 823-854.
Ross, K., 19
Books, Londo
Tao, Q., Le
detection usin
Data. In: P
Convention «
Louisiana, |
Photogramme
364-373.
Tittley, B., S
integration of
for coastal st
Canada: A |
Thematic Cor
Coastal Envir:
LA, pp. 1.225-
UNDP, 1976.
Water Resour
Nations Food
BOT/71/506, :
Wulder, M., ^
Local maximu
and basal area
Sensing of Em
