for improving classification accuracy of remote sensing data. Yuhas, R. H. and A. F. H. Goetz. 1993. Comparision of
In : International Archives of Photogrammetry and Remote airborne (AVIRIS) and spaceborne (TM) imagery data for
Sensing , Washington D. C., USA, Commission VII, pp. 80- discriminating among semi-arid landscape endmembers, In:
86. Proc. of Ninth Thematic Conference on Geologic Remote
Sensing, Pasadena, USA, pp. 503-511.
Table 1. Weighted mean values of the spectral DN and VI influenced by bidirectional effects
Direction Digital Number Vegetation Index
Green Red IR NDVI TVI
North 47.9 39.7 50.9 0.107 0.776
South 60.0 60.4 79.4 0.173 0.797
East 55.7 52.4 73.3 0.156 0.808
West 51.0 45.7 53.1 0.058 0.744
Table 2. Vegetation indices of four different directions in testsite A (mountainous forest) and in testsite B
(reeds vegetation on flat terrain).
Direction Site A (Morundai) Site B (near Shinho)
NDVI TVI NDVI TVI
North 0.049 0.741 -0.099 0.633
South 0.374 0.935 -0.101 0.631
East 0.291 0.889 -0.107 0.627
West -0.211 0.538 -0.102 0.631
Table 3. Normalized Difference Vegetation Index (NDVI) values for dirrernt land cover types on flat terrains.
Reeds Sand dune | Urban | Vinyl(Green) Rock & Fallow & Arable
vegetation house Sands farming
NDVI 0.011 -0.110 -0.097 -0.025 -0.071 0.077
Table 4. Comparison of the SAM classification with GIS supported in-situ analysis in percentage of the
classified of the classified pixels.
Class Number of Pixels Cover % Number of Re.Sample Cover %
Unclassified 111039 9.08 - -
Forest 378379 18.87 39 28.06
Urban 81991 6.70 10 7.19
Winter Agricultural 232493 19.01 16 1151 of
Vinyl(Green)house = - 8 5:75
Barren 6046 0.49 - „11 30
Reeds Vegetation 54168 4.43 5 3.60
River 69881 5.71 5 3.60
Sand dune = - 1 0.72
Coastal sea 436588 35.70 SS 3037 d
Total 1222940 100 139 100
344
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996
