04 
ta 
on 
nd 
m 
nd 
OS. 
n, 
ng 
Sis 
for 
of 
ain 
PE 
AR 
de. 
318, 
RS 
rial 
Int. 
1Z), 
nd- 
| of 
the 
ms. 
and 
atic 
Int. 
en), 
ıbH 
IMAGE PROCESSING AND GIS TOOLS 
FOR FEATURE AND CHANGE EXTRACTION 
Costas Armenakis* and Florin Savopol 
Natural Resources Canada (NRCan) 
Centre for Topographic Information (CTI), Geomatics Canada 
615 Booth Str., Ottawa, Ontario, Canada K1A OE9 
{armenaki, fsavopol} @NRCan.gc.ca 
Commission IV, WG IV/7 
KEY WORDS: Mapping, Acquisition, Extraction, Change, Revision, Semi-Automation 
ABSTRACT: 
Currently the Centre for Topographic Information, Geomatics Canada, NRCan is involved in issue-based programs, such as the 
Geomatics for Northern Development and the Reducing Canada's Vulnerability to Climate Change. As in many mapping 
organizations, the projects within these programs expect the delivery of geospatial data and information in much shorter time periods 
compared to operations in the past due to the external pressures and the availability of new data sources and technology. This 
increasing demand for delivery in shorter time imposes a need for rapid approaches for the extraction of topographic features and the 
detection of landscape changes from imagery. Considering the continuous dwindling of resources, 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. To implement rapid processes for mapping operations, such as feature 
recognition, feature extraction and change detection we have considered the possibilities offered by a) the new kinds of data sources 
and especially the availability of panchromatic and multispectral digital data; and b) the tools and techniques available in image 
processing (IP) and GIS packages respectively and how these tools can be used to accelerate the execution of mapping operations. 
Two case studies, one of which includes the application of CTI's semi-automated change detection approach, are presented to 
demonstrate the potential, applicability and usefulness of this approach. 
1. INTRODUCTION 
The Centre for Topographic Information, Geomatics Canada, 
Natural Resources Canada is involved in  issue-driven 
initiatives, such as the Geomatics for Northern Development 
Program and the Reducing Canada's Vulnerability to Climate 
Change Program. Certain projects under these programs 
conduct acquisition, revision, and monitoring operations for 
spatial data. Consequently, we require to deal with three main 
mapping functions: a) the recognition of features, b) the 
extraction of features, and c) the change detection including 
correction to existing features. All require extensive human 
involvement, as they are time consuming operations. 
Nowadays, the delivery of geospatial data and information is 
expected in much shorter time periods compared to the past. 
This is due to expectations generated by the availability of new 
technology and new data types and sources. Considering the 
continuous dwindling resources (human, budgets) and the vast 
size of the Canadian territory, there is a need to implement rapid 
mapping approaches to reduce both the production time and the 
cost involved. These approaches require not only revisiting of 
the current processes but most important the implementation of 
higher level of automation in the mapping operations. 
Automation for geo-spatial operations, such as feature 
extraction and change detection has been the “pursue of the 
  
* 
Corresponding author 
611 
holy grail" in photogrammetry, remote sensing and spatial 
information sciences. While we may never achieve complete 
automated systems and operations, significant progress has been 
made for various processes under certain conditions and with 
specific data types (e.g., Heipke and Straub, 1999; Baltsavias, 
2004; Zhang, 2004). This has been leading to some automated 
operations but mostly to various semi-automated approaches or 
to tools that can support various semi-automated processes. 
An important factor affecting these operations is the high 
heterogeneity of data and data sources. Data can be vector or 
raster type, their sources could be geodatabases, raster maps, 
airborne and/or spaceborne images with various spatial and 
radiometric resolutions and multi-temporal in nature. Therefore, 
certain processing is required to normalize the data and bring 
them under a common work domain, either at the data level or 
at the information level (Armenakis et al., 2003). In addition, 
higher levels of automation can be achieved when there is 
thematic homogeneity and the operations are feature dependent. 
In this paper we will identify a range of tools and techniques 
available in the functionality of geographic information systems 
(GIS) and the image processing (IP) packages, which can be 
applied usually in combined modes to accelerate the three main 
operations of feature recognition, feature extraction and change 
detection. Two case studies are then presented to demonstrate 
the applicability of several of these tools in mapping operations.