The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008
1379
developed by GeoAnalytika, an affiliate local company of the
author, the desired accuracy and image quality has been
achieved. The final method used was actually a modified
multistep process using IHS and HSL Inversion Method.
5.2
Ortho-Image Production
After creating the pan fused natural colour satellite image, the
next step is image rectification and registration. This stage
involves warping the image to correct for geometric errors and
conform it to a selected map shape in the correct coordinate
system. These five major steps summarizes the whole ortho
rectification process namely selection of map projection and
coordinate system, identification of ground control points
(GCPs) from the image, collection of ground control points with
a GPS, rectification using a suitable mathematical model,
reprojection to a standard map projection and quality control
and assessment.
A total of 33 identified GCPs were successfully identified,
occupied, and established in a local map projection and datum
as required for the final base map. The distribution of points is
shown in Figure 2. In order to meet the accuracy requirements
for the basemap, a high accuracy differential GPS survey was
conducted. Occupation time per control point was set from 15-
20 minutes, which ensures that enough satellite observation data
is acquired to produce a high accuracy position per point.
Results of the Differential GPS survey results were summarized
as follows: Maximum RMS error = 0.017 meters, Average RMS
= 0.009 meters, Maximum Standard Deviation = 0.143 meters,
and Average Standard Deviation = 0.05 meters. The high
accuracy survey results ensures that positional errors will be
kept to a minimum and shall not affect the accuracy of the final
basemap.
After all GCPs have been collected and processed, a
rectification model is then created. The average GCP RMS error
is 0.89 meters. In addition, the RMS error of the GCPs
exhibited an unusually high deviation from the mean error with
differences from a high of 4.78 meters (8 pixels) to a low of
1.13 pixels (0.68 meters) even with repeated checks of their
plotted position in the image. It was deduced that the errors
were inherent in the image because of pre-rectification done by
the image supplier. An alternative method is using a delauney
triangulation. This approach fixes and preserves the image point
locations and uses a linear stretch to warp the pixels in between
the control points. Thus, there is virtually zero (0) error at the
control points and the errors are distributed linearly between
control points. This method assumes that the GCP image points
have been chosen as carefully as possible to match the actual
ground feature with minimum doubt.
The result is a geometrically sound UP Campus base map image
with sub-meter accuracy and visual quality that can rival
conventional color aerial photography suitable for large scale
mapping purposes up to 1:2,000 (Figure 2).
Figure 2. UP Campus Base Map Showing Boundary and
Distribution of GCPs in the Project Area
5.3 Visual Interpretation of High Resolution
QuickBird Imagery
A compilation of land titles was used to define the extent of
university property (Figure 2). The lots are plotted, checked for
traverse errors, and then consolidated to form the whole
coverage of the campus using CAD software. The boundaries
were overlayed on the base map image to determine the extent
of the study area. Remote sensing softwares were used to view
and visually interpret the image to determine locations of areas
occupied or directly utilized by informal settlers.
Ward and Peters (2007) used visual interpretation of high
spatial resolution multispectral IKONOS satellite image to
identify low-income informal homestead subdivisions (IFHS,
also known as colonias) in peri-urban areas of US metropolitan
areas. This process of visual interpretation of high spatial
resolution satellite data is not automated, but requires the
systematic search, identification, and delineation of the target
features by the analyst.
The object of study which is informal settlements is readily
identifiable on the image because of the high resolution and
excellent visual quality of the image. There were two identified
qualified categories of informal settlements for this study. The
first is the “slum” type characterized by MMUSP (2002) to have
lack of spatial pattern, smaller structures, irregular boundary
demarcation, clustering and uneven spread, different reflectance
and locational attributes. These settlements are considered to be
the miserable or depressed areas, normally lacking in basic
services such as electricity, water, and communication lines.
The second type is a “semi-formal” type characterized by a
more decent neighborhood similar to a low-cost housing with
provisions of the basic services and amenities. These
settlements, although not physically informal, lack necessary
tenure or permit to legally occupy the property.
5.4
GIS Mapping
Abbott (2003) has pointed out the value of GIS for the
evaluation of informal settlements. In his study, vegetation
cover and land use have been quantified by visual interpretation
of 1:8000 scaled air photos using the Autocad 14 (AutoDesk)
and ArcView 3.2 (ESRI) softwares. The digitization of the