International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004 
  
* [mage differencing methods assume that the differences 
between the radiometric values are due to changes in the 
object space. Indeed these differences could be a result of 
other factors, such as different atmospheric conditions, 
illumination conditions, changes in soil moisture and 
sunlight angle. Several solutions were suggested to 
overcome such a problem. Basically, these solutions depend 
on image enhancement and radiometric corrections that tend 
to reduce radiometric differences between images under 
consideration. 
= Most of these methods require a decision as to where to 
place the threshold boundaries in order to separate the areas 
of changes from those of no change (Singh, 1989). In fact, 
classical techniques perform thresholding based on empirical 
strategies or manual trial and error procedures, which 
significantly affect the reliability and the accuracy of the 
final change detection results (Li et al., 2002). 
= In general, classification methods require two or more bands 
for the classification process. This is not always available 
especially when dealing with aerial images that represent an 
important source of historical information needed for change 
detection purposes. 
" ]mage  differencing techniques are sensitive to 
misregistration between the reference and input images 
(Singh, 1989; Townshend et al., 1992; Li et al, 2002;). 
Literature pointed out that accuracy of the image registration 
process is the key factor that controls the validity and 
reliability of the change detection outcome. 
In summary, uncertainty in the change detection outcome relies 
on two factors. Firstly, the detected changes might be biased by 
inaccurate — rectification/registration procedure (geometric 
differences). Secondly, it is affected by possible radiometric 
differences due to atmospheric changes and/or different sensor 
types. To overcome the problem of geometric differences, this 
study will investigate and develop a semi-automated, accurate, 
and robust registration paradigm that guarantees accurate co- 
registration which is required for reliable change detection 
(Section 2). To overcome the problem of radiometric 
differences, derived edges from the registered images are used 
as the basis for change detection. The utilization of edges is 
motivated by the fact that they are invariant with respect to 
possible radiometric differences between the images in question 
(Section 3). Section 4 demonstrates the proposed methodology 
of change detection. Experimental results using real data, which 
proves the feasibility and robustness of the suggested 
methodology, are discussed in Section 5. Finally, conclusions 
and recommendations for future work are discussed in 
Section 6. 
2. GEOMETRIC DIFFERENCES 
High resolution overlapping scenes captured by space-borne 
platforms and aerial images are becoming more available at a 
reasonable cost. These images represent the main source of 
recent and historical information that are necessary for change 
detection application. Due to different imaging systems, spatial 
resolutions, viewing points and perspective geometry between 
these temporal images, geometric differences should be 
expected. Reliable change detection is contingent on accurate 
compensation of these differences among the involved images. 
The proposed registration methodology will accurately align the 
images in question regardless of possible geometric differences. 
In general, an image registration process aims at combining 
data and/or information from multiple sensors in order to 
achieve improved accuracies and better inference about the 
environment than could be attained through the use of a single 
sensor. An effective automated image registration methodology 
must deal with four issues (Habib and Al-Ruzouq, 2004); 
namely registration primitives, transformation function, 
similarity measure, and matching strategy. The following 
subsections briefly discuss the rationale regarding these issues. 
2.1 Registration primitives 
To carry out the registration process, a decision has to be made 
regarding the choice of the appropriate primitives (for example, 
distinct points, linear features, or homogeneous regions). In this 
research, straight-line segments are used as the registration 
primitives. This choice is motivated by the following facts: 
= Straight lines are easier to detect than distinct points and 
areal features. Moreover, the correspondence between 
conjugate linear features in the input imagery becomes 
easier. 
= [mages of man-made environments are rich with straight-line 
features. 
= Jt is straightforward to develop mathematical constraints 
(similarity measures) ensuring the correspondence of 
conjugate straight-line segments. 
" Free-form linear features can be represented with sufficient 
accuracy as a sequence of straight-line segments (poly- 
lines). 
After selecting straight-line segments as the registration 
primitives, one has to make a decision regarding on how to 
represent them. In this research, the line segments are 
represented by their end points. This representation is chosen 
since it is capable of representing all line segments in 2-D 
space. Also, it will allow for a straightforward similarity 
measure that mathematically describes the correspondence of 
conjugate line segments. It should be mentioned that the end 
points defining corresponding line segments in the imagery 
need not be conjugate, Figure 1. 
2.2 Registration transformation function 
The second issue in a registration procedure is concerned with 
establishing the transformation function that mathematically 
describes the mapping function between the imagery in 
question. In other words, given a pair of images, reference and 
input images, the transformation function will attempt to 
properly overlay them. Habib and Morgan (2002) showed that 
affine transformation, Equation 1, could be used as the 
registration transformation function for imagery captured by 
satellite imaging systems with narrow angular field of view 
over relatively flat terrain (a terrain with negligible height 
variations compared with the flying height). 
= 
a, a ux 
E + 
/ bil 5 ob y 
Ne 
where 
(x, v): coordinate of a point in the reference image 
Internai 
x,y") 
  
Figure 1. 
2.3 Sir 
The nex 
similari 
necessa 
conjuga 
depend: 
respecti 
primitiv 
points, ° 
Assumi 
corresp« 
1, the « 
fact th. 
corresp: 
transfor 
Such a | 
betweer 
referenc 
input in 
mathem 
points © 
where 
(p,0) 
(x; ’ AH ) 
Anothei 
point 2 
2.4 Mi 
To aut 
controll 
measur 
framew