The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008
Figure 6 shows ED and MEAN statistics of nine land cover
types. Generally speaking, bare ground shows the highest
MEAN value, while water bodies show the lowest. In between
there are numerous ‘confusing’ classes including three types of
residential area, farmland, grassland and woodland. It is
obvious that the MEAN alone does not provide enough variance
to distinguish residential areas from their background. Old
urban and rural residential areas show greater ED values as
their buildings are assembled densely. Roads and cars on the
road have clear edges resulting in a high ED value. New urban
residential areas have sparse buildings, thus their ED values are
smaller than the other two types of residential areas. Trees in
woodland often show clear crown and shadows on high-
resolution images, thus is more likely to detect edge point with
this cover type. Other background classes such as water,
grassland, bare farmland and bare ground have an ED value
close to zero. Thus, it is clear that ED is a good candidate for
the detection of residential areas with the background except
road and wood land.
Figure 7 presents the contrast group to distinguish residential
areas and background. New urban residential area, road, rural
residential area, old urban residential area and wood land show
more contrast than water, grassland, bare farmland and bare
ground.
Figure 8 shows orderliness group in describing pixel variation.
All three kinds of residential areas have less order or uniform
surface: woodland and road show medium uniform surface,
while water, grassland, bare farmland and bare ground represent
a uniform surface.
Figure 9 uses SD and COR to describe differences of
residential areas and their background. It can be seen that new
built urban residential area, old urban residential area, rural
residential area, road and woodland all show greater SD values,
which means that the pixel DN values are more diversified. In
contrast, bare ground, grassland, water and barren farmland all
show smaller SD values. All three types of residential areas are
characterized with smaller COR values than the background
classes, suggesting more discrete and less dependent spectral
distribution.
4. CONCLUSIONS
In this paper, we investigated methods for selecting and
evaluating texture parameters (window size, quantization level,
displacement and orientation) for the identification of
residential areas based on JM-distance. grey level co
occurrence matrix (GLCM) and edge density (ED) approaches
with candidate nine texture measurements (contrast,
homogeneity, dissimilarity entropy, energy, mean, standard
deviation, correlation and edge density) is selected as candidate
texture measurements. The texture parameters are selected
based on Jeffries-Matusita distance (JM distance) between
residential and its background in corresponding texture space.
IKONOS panchromatic imagery has been used as example and
the optimal texture parameters were selected by using the
proposed method. Further studies will be focused on the
selection of optimal texture combination to improve the
residential classification results.
ACKNOWLEDGEMENT
This project is financed by two items of the National Natural
Science Foundation of China (Contract No. 40337055 and
Contract No. 40501062).
REFERENCES
Anys, H. and He D.C., 1995. Evaluation of textural and
multipolarization radar features for crop classification, IEEE
Transactions of Geoscience and Remote Sensing, 33(5): 1170-
1181.
Cobum, C.A., Roberts, A.C.B., 2004. A multiscale texture
analysis procedure for improved forest stand classification,
International Journal of Remote Sensing, 20 October, 25(20),
4287-4308.
Conners, R.W and Harlow, C.A.,1978. Equal probability
quantizing and texture analysis of radiographic images,
Computer Graphics Image Processing, 8, 447-463.
Canny J.F. 1986. A computational approach to edge detection.
IEEE Transactions on Pattern Analysis and Machine
Intelligence, 8(6): 679-698.
Gong, P., 1991. A comparison of spatial feature extraction
algorithms for land-use classification with SPOT HRV Data,
Remote Sensing of Environment, 40, 137-151.
Haralick, R.M. 1979. Statistical and structural approaches to
texture, Proceedings of the IEEE, 67, 786-804.
Karathanassi, V., Iossifidis, CH., and Rokos, D., 2000. A
texture-based classification method for classifying built areas
according to their density, International Journal of Remote
Sensing, 21(9), 1807-1823.
Myint, S.W. and Lam, N., 2005. A study of lacunarity-based
texture analysis approaches to improve urban image
classification, Computer, Environment and Urban Systems, 29,
501-523.
Shaban, M.,A., and Dikshit, O., 2001. Improvement of
classification in urban areas by the use of textural features: the
case study of Lucknow city, Uttar Pradesh, International
Journal of Remote Sensing, 22(4), 565-593.
Soh, L.K. and Tsatsoulis, C., 1999. Texture analysis of SAR sea
ice imagery using grey level co-occurrence matrices, IEEE
Transactions on Geoscience and Remote Sensing, 37(2), 780-
795.
Swain, P.H. and Davis S.M.,1978. Remote Sensing: The
quantitative approach, New York: McGraw-Hill,