clarify the ambiguous areas. A comparison of 
the data collected from the false color image 
using bands 4, 3 and 2, to data collected from 
the true color image using bands 7, 2 and 1, 
showed that the operators obtained better results 
from the true color image. 
3.2.3 Quality Control 
Plots of the original derived file (before the 
features were masked) and of the data collected 
from Landsat were produced at the 1:50 000 
scale. Since both the derived data and the data 
collected from the Landsat image were 
registered using the same control, the 
comparison to check their geometric fit was 
made visually by overlaying the plots. It was 
found that the collected Landsat data was jagged 
due to the point to point collection method used. 
The Douglas-Peucker filter and a smoothing 
operation were applied interactively to the 
Landsat data, which brought the two files to a 
closer geometric fit. The maximum spatial 
discrepancy between the two data sets existed in 
gravel pits and forest areas. These are two 
types of land uses that have the potential to need 
frequent updating and are not likely errors but 
valid changes. Linear features and water 
boundaries were found to deviate from their 
position in the derived file often by less than 50 
meters. When the plots were done at the 
1:250 000 scale the linear feature discrepancies 
are almost negligible. It was found that the 
collected Landsat data matched the derived data 
within the accuracy and tolerance required for 
the 1:250 000 map. 
3.2.4 Production of Paper Maps 
The capabilities of CARIS for the integrated 
vector/raster data output for the production of 
image maps were also evaluated. Cartographic 
editing, placing toponymic data and contour 
labels, generating the border, UTM grid and 
map surround, for the production of line maps 
is performed by the CARIS cartographic editing 
system. The cartographic vector data was 
exported to POSTSCRIPT format for input to 
Products and Services Division's, (PSD), 
SCITEX system. The co-registered raster data 
was exported in PIP (PCI-BSQ intermediate file 
format) for input to PSD's PCI system, where it 
was prepared for plotting. To achieve good 
visually accepted image print, the TM image 
was resampled using cubic convolution with 
output pixel size of 25m. Image maps at 
1:50 000 and 1:100 000 scales were produced 
on the IRIS plotter and were composed of the 
Landsat TM image, contours from the vector 
file, and the border and grid. 
4.0 REVISION OF 1:50 000 NTS MAPS 
An experiment was also performed to evaluate 
the potential of updating the 1:50 000 
topographic maps from digitized photography. 
Both digital rectified and orthorectified 
photographs were used for the test. The 
1:50 000 data set had been stereo compiled from 
1984 photography. This data set contained 
unstructured positional data as it was compiled, 
not the cartographically edited data as presented 
on the map. The vector data files in SIF format 
were loaded and converted into the CARIS data 
files required for processing and display. Aerial 
photography at scale of 1:50 000 was used to 
evaluate the accuracy of the digital revision 
process. The Helava DPW770 system was 
used for the scanning and the production of 
digital orthophotos. The diapositives were 
scanned at 25 microns (1000 dpi). A 50m DEM 
grid was generated and used to create the 
orthophotos. The image was resampled with 
nearest neighbour with an output pixel size of 
2m ground resolution (about 40microns). Both 
digitized photography and othophotography 
were exported in DOS TIFF format and using 
the CARIS data exchange utilities were 
converted to CARIS IPV raster format. 
4.1 Measurement and Evaluation of Collected 
Data 
The number and location of the control points 
were based upon the points used for the 
aerotriangulation, which covered the extent of 
the digital photography. For registering the 
orthophotograph to the "old" map digital data, 
an affine transformation was used. The 
digitized photograph was rectified using the 
projective transformation. A total of nine (9) 
control points were used for the 
transformations. With the raster and vector 
images combined, the coordinates of thirty three 
(33) check points were measured in the raster 
image. These points were evenly distributed 
throughout the registered area and were well 
identifiable. Their ground coordinates were 
compared with the "true" values of the existing 
positional data. For the rectified photograph the 
standard deviation of the differences was 
+12.5m in x and +16.0m in y. For the 
orthophoto the standard deviation of the 
differences were +2.9m in x and +4.9m in y. 
292 
  
The r 
NATC 
(CMA 
equive 
of +8. 
Althoi 
these s 
the di 
good | 
visual 
than tl 
Suppo! 
capabi 
As a 1 
CCM' 
genere 
raster : 
Havin 
area t 
data se 
to pro 
order 
1:50 
1:250 
site we 
the acc 
some 
Lands 
revisic 
uncert; 
feature 
to im 
(aeria. 
checks 
image 
combi 
produc 
CARI: 
integré 
spatial 
It can 
produc 
produc 
more 
integre 
from c 
with t] 
sheet. 
The cc 
Auger