International Archives of Photogrammetry and Remote Sensing, Vol. 32, Part 7-4-3 W6, Valladolid, Spain, 3-4 June, 1999 
165 
be very effectively determined in the IRS colour composite, 
whilst subclasses are only recognisable in individual instances. 
This stresses the value of the IRS data for updating of biotope 
maps and significant reduction of the needed fieldwork. The 
issue of DTMs was addressed repeatedly, these being crucial for 
the identification of erosion hazards, assessing flood risks, as 
well as mapping of urban climate. 
For the automatic identification of urban structure types, the 
automatic classification of urban structure types, in particular of 
residential areas, the results of the morphologic analysis 
(chapter 4.2), as well as calculated texture images (skewness, 
texture measures, etc.) are overlaid with the boundaries of the 
building-block maps, in order to calculate correlation 
coefficients. This allows a statistical comparison between the 
urban structure types evaluated by visual classification (as 
described at the beginning of this section) and the mean values 
of structural and textural features for the same block areas. The 
multispectral classification results are not used in this process, 
because they do not provide a differentiation of structure types, 
but just a global class of "built-up areas". Using the calculated 
correlation coefficients, threshold values can be determined by 
visual estimation, which enable a partly automatic classification 
of residential areas. First results show that in particular by 
overlaying structural/morphologic parameters different 
residential areas can be differentiated. These investigations will 
be continued for a qualitative improvement. 
6. CONCLUSIONS AND OUTLOOK 
The results of this paper demonstrate that fused IRS-1C data 
constitute a very good tool, especially for regional planning 
with its medium working scales. The application areas of this 
data could be considerably extended, if the data are suitably 
processed, as it is demonstrated in this investigation by firstly 
combining high resolution panchromatic images with 
multispectral ones followed by a hierarchical classification 
scheme including spectral and morphological analysis methods. 
Fusion methods are being checked especially to ensure the 
optimal preservation of the spectral and the geometri 
cal-structural information in the fusion product, which is a 
prerequisite for the following processing. Visual and statistical 
comparisons indicate that fusion methods using local filter 
techniques provide more stable results (regarding the 
preservation of the geometrical characteristics (HPF) as well as 
the spectral characteristics (LMM and LMVM)). 
Paying more attention to the morphology and spatial patterns of 
settlement areas, a hierarchical classification scheme, which 
combines spectral and morphological analysis methods, is 
described and tested. These methods result in significantly 
improved classification accuracy (evaluated visually) compared 
to a conventional multispectral classification procedure. The 
two goals this project has been aiming at, i.e. the precise 
identification of urban (settlement) areas and the urban 
structure-type classification were mostly reached. However, the 
quality of the results can be improved by future work. 
Especially, the potential use of IRS-1C data for the automatic 
classification of urban structure types has to be investigated 
more intensively. 
ACKNOWLEDGEMENTS 
The work described in this paper is part of a research project 
funded by the German Research Foundation (DFG). 
Principal 
structure type 
Structure type 
Classification characteristics 
Built-up areas 
High-density built-up areas 
misclassification as densely built-up areas (in spots) 
Densely built-up areas 
misclassification as terrace houses or high-density built-up areas (in spots) 
Terrace housing 
misclassification as densely built-up areas (in spots) 
One-family housing 
characteristic rough texture 
Large-scale commercial areas 
features: large coherent areas, suburban location 
Green areas 
Parks and green spaces 
single and groups of trees with characteristic signatures 
Forests 
characteristic spectral signatures 
Cemeteries 
mid to fine texture, in spots misclassification as allotments 
Allotments 
fine texture, in spots misclassification as cemeteries 
Sport facilities 
characteristic area outlines 
Agricultural areas 
Pasture land 
large coherent areas, characteristic spectral signatures 
Arable land 
large coherent areas, in spots misclassification as pasture land 
Traffic areas 
Streets, squares 
in spots occluded by buildings, not necessary for block construction 
classification 
Railway areas 
characteristic area outlines 
Water 
clear, no problems 
Excavation sites 
in spots misclassification as construction sites 
Construction sites 
characteristic spectral signatures and area outlines 
Table 2. Classification scheme of urban structure types.