e 
d 
this case, is to determine whether a portion of a specific vector falls 
within a particular pixel in a raster grid. Generally, when one needs 
to integrate vector data into an image analysis environment, the 
output image size and coordinates are set. That is, the size and 
complexity of the output is fixed to within a certain number of pixels 
and lines. 
A visual comparison of the results are shown in Table 1 and Table 
2. Overall, it was found that bringing a rasterized vector coverage 
into the raster domain was preferable, in terms of general 
appearance. Thatis, the rasterized vectors had the greatest visual 
appeal when overlayed on an image. In the raster domain, one is 
bound by the fact that sub-pixel registration is not considered, and 
that where a vector lies within a pixel is academic. The 
performance of this method is summarized in Table 1. 
When one looks at a vector representation of extracted features, it 
becomes evident that the edges of the vectors often do not match 
up. This is largely dependent on the quality of the geographic 
referencing of the original raster image and the effects of any edge 
smoothing that was performed on the boundaries. The overall 
performance of this method is summarized in Table 2. 
5.0 CONCLUSION 
Automated feature extraction techniques can not replace manual 
digitizing, as of yet. The potential for image segmentation or 
feature extraction to supplement the job of an operator is certainly 
there. Automated and semi-automated techniques are desirable to 
enhance operational turn-around time for getting data through a 
system. The logical end of this process is a more efficient system 
for decision-making. As with most things, there are strong 
elements of give and take, in this case, with respect to image 
analysis and Geographic Information Systems. The more 
complicated the data, the greater the demands on the operator to 
manage the data. Data abstraction is an important aspect of 
decision making tools, but the user must always keep in mind the 
accuracy of the information, and hence utility and value of the 
decisions made. 
369 
REFERENCES 
Estes, J.E. and D.S. Simonett, eds. Manual of Remote Sensing: 
Volume II (Falls Church: American Society of Photogrammetry, 
1974) pp. 869-887. 
Manore, M., R.J. Brown, K. Korporal and H. Press (1989) "GIS in 
the Analysis of Satellite Data for Vegetation Monitoring: The Crop 
Information System" Proceedings of the National Conference on 
Geographic Information Systems - GIS 89 pp. 118-124. 
O'Brien, D (1991) "Computer Assisted Feature Extraction 
(INTEREX)" Proceedings of the 14th Canadian Symposium on 
Remote Sensing pp. 423-429. 
Schowengerdt, R.A. and R.A. Pries (1988) "interactive Image 
Feature Compilation for Geographic Information Systems" SPIE - 
Recent Advances is Sensors, Radiometry, and Data Processing for 
Remote Sensing Vol. 924. pp. 305-311. 
Taylor, T., P. Landriau and A. Scott (1991) "Context Classification 
for Road Network Extraction from Landsat and SPOT Imagery" 
Proceedings of the 14th Canadian Symposium on Remote Sensing 
pp. 413-419. 
Van Cleynenbreugel, J., F. Fierens, P. Suetens and A. Oosterlinck 
(1990) "Delineating Road Structures on Satellite Imagery by a GIS- 
Guided Technique" Photogrammetric Engineering and Remote 
Sensing Vol. 56, No. 6. pp. 893-898. 
Walsh, Stephen J., D.R. Lightfoot and D.R. Butler (1987) 
"Recognition and Assessment of Error in Geographic Information 
Systems" Photogrammetric Engineering and Remote Sensing Vol. 
53, No. 10, pp. 1423-1430. 
Wang, M. and P.J. Howarth (1991) "Some Generic Issues in 
Spatial Data Integration” Proceedings of the 14th Canadian 
Symposium on Remote Sensing pp. 444-447. 
Zelek, J.S. (1990) "Computer-Aided Linear Planimetric Feature 
Extraction” (1990) IEEE Transactions on Geoscience and Remote 
Sensing Vol. 28, No. 4. pp. 567-572.