AUG 
i 
A 
| 
  
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part Bl. Istanbul 2004 
  
After the classification had been completed, the next stage was 
to evaluate imagery for its detail to allow counts of individual 
structures. The ability to count individual structures allows 
quantification of urban change, into the number of households, 
on which the demarcation of EA's is based. 
This was done by attempting house counts for each individual 
structure class listed above. Two polygons from each class 
were overlayed on each of the 5 different sets of Spots 
resolution imagery. House counts were generated on each 
resolution set by physically attempting to place a point on each 
visible structure. These were all compared to a base data set 
generated from 1:10 000 aerial photography at 1m resolution. 
3.4 Results 
The results have shown that Spot 5 has an advantage for census 
applications compared to lower resolutions, especially when 
attempting to classify or quantify urban change. Imagery was 
evaluated in three stages for its use, namely, urban change 
detection, urban structural classification and detail of imagery 
to allow counts of buildings. 
During the mapping of change as described above, all the 
potential areas of change in the Pretoria areas could be detected 
by all the sensors from Spot5. It has to be kept in mind that this 
study has been done in a metropolitan area with clear urban 
structures, which is easier to identify and recognise, compared 
to rural settlement patterns. It is however expected not to be the 
case in all urban areas and patterns. Based on previous 
experience of similar mapping projects, this could be the case 
in some rural areas, where villages with low densities of small 
structures occur. 
The classification of the change area according to the urban 
structural classes indicated that there is a differentiation on this 
factor between the Spot 5 sensors. Although some structures 
could be recognised and identified from all the sensors, it was 
not possible to classify all structures. Spot 5 Colour (10m) and 
Panchromatic (5m) could classify all the classes except the low 
cost housing and informal settlements, which are classes of 
dense, but very small houses. The Spot 5 merge of Colour 
(10m) and Panchromatic (5m), Colour (10m) merge with 
supermode (2.5m) and supermode (2,5m) could clearly classify 
all classes. 
The final stage to perform building counts in order to quantify 
urban growth indicated that only Spot Colour (10m) merge 
with supermode (2.5m) and supermode (2.5m) could 
distinguish buildings in all structural classes, except low cost 
housing and informal settlements. This result has created new 
potential to reduce cost and time of Census Cartography as 
performed in preparation for Population and Housing Census 
surveys. It is now possible to use only one set of imagery to 
perform three procedures, while previously it required two or 
three data sets and the same amount of workflow 
Spot 5 regular updates which allow an annual recording of 
specific metropolitan areas, adds the ability to generate 
historical records of spatial patterns and growth occurring 
around the metropolitan perimetre. Information on previous 
years can be useful to analyse growth patterns and serve as a 
base for future modelling. 
4. CONCLUSION 
The possibility to reduce the amount of image data sets from 
potentially two to only one, as well as the subsequent reduction 
in manpower required for processing and mapping creates the 
following advantages for Census Cartography: 
* Only one set of imagery as input to the census 
cartography process 
* Only one workflow procedure to perform with less 
cost and manpower 
* Closer date of acquisition to census date 
* Uniform image coverage for areas where population 
are concentrated 
* Building counts opens potential for inter-census 
updates 
The potential to adapt this model to other parts of the world 
creates the potential to apply this approach to any region with 
rapid urban growth patterns. 
5. ACKNOWLEDGEMENTS 
Spot Image for supplying the SpotS imagery for this study, as 
listed in this paper. 
554