International Archives of Photogrammetry and Remote Sensing, Vol. 32, Part 7-4-3 W6, Valladolid, Spain, 3-4 June, 1999
181
Landcover and
Landuse Classes
Classification based
on Spectral bands
Classifications based on Spectral
and Textural bands from matrix 1
Classifications based on Spectral
and Textural bands from matrix 2
Crop 1
Crop 2
Trees
Grass
Parks
Apartment-block areas
Low density urban areas
High density urban areas
Water
cl
c2
c3
c4
c5
c6
c7
c8
c9
90
82
86
90
82
88
96
88
88
78
80
84
84
84
90
96
92
82
98
90
96
92
90
94
96
90
92
98
88
94
94
92
96
94
94
80
68
76
80
85
78
80
92
94
78
68
72
76
78
74
84
96
96
84
64
68
72
78
72
86
90
94
86
62
70
74
76
74
78
92
94
90
98
90
94
92
90
98
96
94
80
Total
80.4
82
84
84.2
81.7
88.2
94.2
92.9
84.4
Table 3. Classification accuracy (%) for different classification methods and window sizes.
4. CONCLUSIONS
The classification accuracy with the new procedure (termed
procedure 3) is 94.2%, 10% higher than the accuracy achieved
using procedures 1 and 2. This reveals that textural features
derived from multispectral images are a very valuable source of
spatial information and an important clue for landcover and
landuse classification. The grey level co-occurrence matrix for
textural measure calculations is an important factor, which
affects the fidelity of textural features. The textural measures
based on matrix 2 can more effectively reveal spatial forms of
landcover and landuse types in multispectral images.
An inappropriate window size can reduce the classification
accuracy, and the window sizes of 5 x 5 and 7x7 can be
considered, based on the performed tests, as the appropriate
ones for this set of landcover and landuse categories.
The new procedure is particularly suitable for classification of
images containing complex spectral components, like urban
regions.
REFERENCES
Baraldi A. and Parmiggiani F., 1990. Urban area classification
by multispectral SPOT images. IEEE Transactions on
Geoscience and Remote Sensing, 28(4): 674-679.
Benediktsson E.B., Swain P.H. and Ersoy O.K., 1990. Neural
network approaches versus statistical methods in classification
of multi-source remote sensing data. IEEE Transactions on
Geoscience and Remote Sensing, 28(4): 550-552.
Bischof H., Schneider W. and Pinz A.J., 1992. Multispectral
classification of Landsat images using neural networks. IEEE
Transactions on Geoscience and Remote Sensing, 30(3): 482-
490.
Fausett L., 1994. Fundamentals of Neural Networks -
Architectures, Algorithms, and Applications. Prentice Hall
Intemational.Inc., London.
Fung T. and Chan K., 1994. Spatial composition of spectral
classes: A structural approach for images analysis of
heterogeneous land-use and land-cover types. Photogrammetric
Engineering & Remote and Sensing, 60(2): 173-180.
Gong P. and Howarth P., 1992. Frequency-based contextual
classification and grey-level vector reduction for landuse
identification. Photogrammetric Engineering & Remote
Sensing, 58(4): 423-437.
Haralick R.M., Shanmugam K. and Dinstein I., 1973. Textural
features for image classification. IEEE Transactions on System,
Man, and Cybernetics, 3(6): 610-621.
Jensen J.R., 1979. Spectral and textural features to classify
elusive landcover at the urban fringe. Professional Geographer,
31(4): 400-409.
Marceau D.J., Howarth P., Duboise J. and Gratton D., 1990.
Evaluation of the grey-level co-occurrence matrix method for
land-cover classification using SPOT imagery. IEEE
Transactions on Geoscience and Remote Sensing, 28(4): 513-
519.
Paola J.D. and Schowengerdt R.A., 1997. The effect of neural
network structure on a multispectral land-use/land-cover
classification. Photogrammetric Engineering & Remote
Sensing, 63(5): 535-544.
Sadler G.J., Barseley M.J. and Barr S.L, 1991. Information
extraction from remotely sensed images for urban land analysis.
Proceedings of European Conference on Geographical
Information Systems, pp. 955-964.
Wharton S.W., 1982. A context-based land-use classification
algorithm for high resolution remotely sensed data. Journal of
Applied Photographic Engineering, 8(1): 46-5.
