inbul 2004 
SEMI-AUTOMATIC REGISTRATION AND CHANGE DETECTION USING MULTI- 
SOURCE IMAGERY WITH VARYING GEOMETRIC AND RADIOMETRIC 
PROPERTIES 
A. F. Habib, R. I. Al-Ruzouq?, C. J. Kim* 
“ Department of Geomatics Engineering, University of Calgary 
2500, University Drive NW, Calgary AB T2N 1N4 Canada — (Habib, Al-ruzouq)@geomatics.ucalgary.ca, 
Cjkim@ucalgary.ca 
PS ICWG IVIV Change Detection and Updating for Geodatabase 
KEY WORDS: Change Detection, Automation, Registration, Feature, Matching, Multi-resolution. 
ABSTRACT: 
Change detection is the process of identifying differences in the state of objects and/or phenomena under consideration by observing 
them at different times. Change detection is important for monitoring and managing natural resources, urban development, 
environmental changes, and disaster assessments. Recent advances in satellite imagery, in terms of improved spatial and temporal 
resolutions, allow for reliable identification and prediction of change patterns. The quality of the image registration process of the 
involved imagery is the key factor that dictates the validity and the reliability of the change detection outcome. The fact that change 
detection analysis might involve multi-spectral, multi-source, and/or multi-resolution imagery captured at different times calls for 
the development of a robust registration procedure that can handle these types of imageries. This paper introduces a new approach 
for semi-automatic image registration using linear features, which can be reliably extracted from imagery with significantly different 
geometric and radiometric properties. The Modified Iterated Hough Transform (MIHT) is used as the matching strategy for 
automatically deriving an estimate of the parameters involved in the transformation function relating the images to be registered as 
well as the correspondence between conjugate lines. Once the registration problem has been solved, the suggested methodology 
proceeds by detecting changes between the registered imagery. Traditional change detection methodologies, which are based on the 
subtraction of intensity images, usually fail due to different illumination conditions, sensors, and/or viewing perspectives at the 
moments of exposure. To overcome these problems, features that are invariant to changes in the illumination conditions can be used. 
Based on this reasoning, derived edges from the registered images are used as the basis for change detection in this research. 
Experimental results using real data proved the feasibility of the suggested approach for deriving a quantitative estimate of changes 
among the registered images. 
1. INTRODUCTION order to derive a suitable training set for the learning process of 
the classifier. The latter performs change detection directly by 
The demand for up-to-date geographic data is increasing due to comparing two images under consideration, without relying on 
fast changes in the real world that are taking place as a result of any additional information (Bruzzone and Prieto, 2000). 
nature and/or human actions. Such changes have to be 
accurately and reliably inventoried to fully understand the 
physical and human processes at work (Estes, 1992). Change 
detection is the process of identifying differences in the state of 
an object or phenomenon by observing it at different times 
(Singh, 1989). It involves the ability to quantify changes using 
multi-resolution, multi-spectral, and/or multi-source imagery 
captured at different epochs. Traditional change detection 
Studies are based on visual/manual comparison of temporal 
datasets (such as satellite scenes, aerial images, maps, etc.). 
However, the huge flux of imagery that is being captured by an 
ever increasing number of earth observing satellites necessitates 
the development of reliable and fast change detection 
techniques. Such techniques are essential to reduce the high 
cost associated with spatial data updating activities. 
In supervised classification, data from two images are 
separately classified, thus the problem of normalizing such data 
for atmospheric and sensor differences between two different 
times is minimized (Singh, 1989). The supervised approach 
exhibits some advantages over the unsupervised, mainly the 
capability to recognize the kinds of land cover transition that 
have occurred, robustness to different atmospheric and light 
conditions at the two acquisition times, and the ability to 
process multi-sensor/multi-source images (Bruzzone and 
Serpico, 1997). A major drawback of the supervised 
classification is that the generation of an appropriate multi- 
temporal ground truth is usually a difficult and expensive task; 
in addition, greater computational and labelling efforts are 
required. On the other hand, unsupervised classification 
primarily entails the creation of “difference images”. It involves 
Several change detection methods have been developed and image differencing, image ratio, vegetation index differencing, 
reported in the literature (Singh, 1989; Townshend et al., 1992; image regressions, change vector analysis (CVA), and principal 
Dowman, 1998; Bruzzone and Prieto, 2000; and Li et al., 2002). ^ component analysis (PCA). Changes are then identified through 
Basically, two main solutions for the change detection problem analysis of (e.g., thresholding) the difference images. 
have been proposed: supervised and unsupervised approaches. 
The former is based on supervised classification methods, 
Which require the availability of multi-temporal ground truth in 
Based on the literature of change detection techniques, the 
following issues have to be considered: 
445