)04 
ter 
hat 
ain 
tial 
  
  
  
sed 
an 
ing 
of 
ind 
Fig 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004 
I). An ISODATA unsupervised classification was applied using 
bands 4,5,7. The class related to water bodies was extracted 
(Fig. 2, in green) and was intersected with the thresholding 
output via an AND Bolean operator to determine the common 
areas between this two water related areas (Fig. 3). 
An area filter then applied to eliminate water body polygons 
smaller than the minimum area size. Following, the slope angles 
were computed from the available DEM and the slope range 0-4 
degrees was extracted (Fig 4) and used as spatial constrain, 
considering that the water bodies are located within this range 
of slopes. Finally the common areas between the extracted 
water bodies and the selected slopes were estimated to 
determine the final areas of water bodies (Fig 5). The 
commission errors (extraction water body areas where there are 
not) and omission errors (omission existing water body areas) 
for the water bodies areas are shown in Figure 6. 
4. CASE STUDY II: COASTLINE CHANGE 
DETECTION 
Monitoring of coastal changes contributes to the development 
of various types of assessments (e.g., impacts, sensitivity, 
vulnerability, erosion hazard) due to climate change including 
changes in the sea-level. One of the projects of the ESS 
Program Reducing Canada’s Vulnerability to Climate Change 
investigates the ocean vulnerabilities to climate change on a 
regional and local scales to provide critical geoscience data to 
other government departments. This information can be used in 
assessing climate change impacts and developing adaptation 
options. The task of this coastal activity is to develop sensitivity 
and impacts assessments in coastal areas of the Arctic and East 
coasts of Canada by using earth observation data to provide the 
necessary spatio-temporal data infrastructure for coastal areas 
and by developing methodologies for feature extraction and 
change detection using image data. The study area is the Arctic 
south-west coast of Banks Island where the settlement of Sachs 
Harbour is also located. The time-series data used were 1961 
aerial photography and 2002 IKONOS imagery. 
The Canny edge detector was used on the 1961 orthophoto to 
semi-automated extract the coastline (Fig. 7), while due to 
systems limitations heads-up digitization was used to extract the 
coastline from the 2002 IKONOS imagery. 
  
  
Figure 7: Extraction of coastline using the Canny edge detector. 
615 
The extracted pixel edges were converted to vector lines via 
R=>V process using snapping, smoothing and pseudo nodes 
removal operations. 
For the estimation of the planimetric change detection of the 
coastline the above described semi-automated feature-based 
approach, developed at CTI and implemented in the ArcGIS 
environment, was used. The changes were determined as 
additions (gains) or deletions (loss) to the land (Fig. 8). 
Preliminary results for approximately 39 km length of coast-line 
show that sea has gained about 446 310 m? and land has gained 
about 351 994 m°, thus resulting in total loss of land of about 
93 316 m° between 1961-2002. 
  
Figure 8: Feature-based coastline change detection between 
1961-2002 (land gains in yellow; land losses in light green). 
5. CONCLUDING REMARKS 
There is a variety of systems and tools available and the trend to 
integrate photogrammetric, image analysis and GIS 
functionalities continues. There are new kinds of data sources, 
such as high resolution images, simultaneous availability of 
panchromatic and multispectral images, and LIDAR data. And 
there are new kinds of requirements and applications that 
required rapid and enhanced mapping operations. 
The implementation of higher level of automation in the 
mapping operations is highly desirable to reduce both the 
production time and the cost involved, especially when dealing 
with the vast size of the Canadian territory and the continuous 
dwindling of resources. To implement rapid processes for 
mapping operations, such as feature recognition, feature 
extraction and change detection we have considered several 
tools and techniques from existing image processing and GIS 
packages which may be used to either enhance the operations or 
as alternatives approaches to the procedures. The integrated use 
of these tools allows also for the development of techniques, 
such as the presented feature-based change detection, which 
significantly improves certain operations and results. 
However, there is no general solution for the various operations 
and presently different tools can be used for different data types 
and occasions. That is, the tools are selected based on the data 
source and the required outputs, while the accuracy, reliability 
and completeness of the results may vary from one application 
to another. This non-standardization of approaches requires 
specialised personnel and continuous training. 
  
  
  
N M