International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004 
  
e.g. 15 cm. The grid points are called as groundel. All the 
groundel in each horizontal plane will be projected back to the 
related aerial images by means of collinearity condition to 
produce the possible orthoimage of this terrain area at some 
height. Therefore, several orthorectificated image patches can 
be generated from several related aerial images. Those 
orthorectificated image patches constitutes a candidate set. 
Therefore, for possible height information, several candidate 
sets of orthoimage patches for this terrain area will be generated 
from those horizontal planes. 
Aerial Image 2 
    
  
   
Aerial Image 1 
*«rial Image j 
  
  
Terrain Area 
Orthorectification 
By using possible height from H1 to Hk 
The optimal set 44 thoimage patches 
by minimal radiometric difference 
Unique Orthoimage patch by Image Fusion 
; 
  
Figure 1 Study Methodology and Flowchart 
As Fig. 1 shown, each height Hi will produce a candidate set of 
orthorectified image patches for a terrain area. By evaluating 
the radiometric difference information among orthoimage 
patches in each candidate set, the optimal set of orthoimage 
patches at some height will be decided. In other words, a 
procedure to decide corresponding height information of this 
876 
terrain area will be developed. Section 2-2 will describe this 
elaborate process. 
After the optimal set of orthoimage patches is decided, an 
image fusion approach will be developed to integrate the 
multiple image radiometric information into unique 
orthorecitifed image patch. Data snooping method [Baarda, 
1968; Wolf and Ghilani, 1997] is often used in the adjustment 
for blunder detection. This method will be employed to exclude 
the inappropriate radiometric information in the optimal set of 
orthorectificated image patches. Afterwards, a simple data 
fusion method is used to integrate the relevant radiometric 
information form multiple image patches into a unique 
orthorectificated image patch. Subsection 2-3 will described the 
method in more details. 
The relevant assessment of accuracy, especially geometric 
accuracies, is also conducted in this paper. Section 3 will 
discuss the relevant problems in accuracies. Experiments and 
discussions will be presented in Section 4. The short occlusion 
will be given in final section. Next subsection will discuss the 
proposed concept in this method in advance. It is called 
“floating plane”. The concept of “floating plane“ will be 
compared with the “Floating Mark” employed in sterescopical 
view photogrammetry. 
2.1 Floating Plane 
In this study, after the approximation height range of a terrain 
area is known, several horizontal planes corresponding to this 
terrain area will be produced according to the necessary height 
accuracy. In photogrammetry, if we imagine that the operator 
stereoscopically measures terrain characteristic points or 
features by using photogrammetric equipments. The operator 
will move floating mark up and down vertically to obtain the 
accurate height information. The idea used in this study is likely 
to the concept of “Floating Mark”. The concept of “Floating 
Mark” is expanded to the concept of “Floating Plane”. 
Therefore, if the algorithm is developed and used to find the 
optimal orthorectificated image patch set at some height range 
is just like the operator to manipulate the “Floating Plane” to 
locate the accurate plane location according to the provided 
radiometric information among multiple images. Similar 
concept of “Floating Plane” can be found in [Collins, 1996]. 
2.2 The Optimal Set of Orthoimage Patches by Minimal 
Radiometric Difference 
As shown in Fig.1, all the groundels in each horizontal plane 
will be projected back to the related aerial images by means of 
collinearity condition to produce one candidate set of 
orthoimage patches for selected terrain area. A procedure must 
be developed to decide the height information for this terrain 
area among all candidate sets of orthoimage patches. 
Firstly, the distances between the central location of this terrain 
area and the relevant aerial images locations (X,Y,Z) are 
calculated and used to decide the nearest aerial image. The 
orthoimage patch generated from this nearest aerial image is 
called as the major orthoimage patch. Then, for each candidate 
set of orthoimage patches, the sum of absolute value of 
radiometric difference information between other orthoimage 
patches and this major orthoimage patch are calculated. Among 
the absolute value of radiometric differences from all candidate 
sets, the minimal is selected as the optimal set of orthoimage 
patches. Namely, the height information is decided in the 
meantime while this optimal orthoimage patch set is decided. 
Inte 
2.3 
Wi 
can 
inf 
As 
at 
san 
ma 
ort] 
bet 
shc 
cor 
int 
als 
Ac 
res 
foll 
wh 
No 
EXC 
of : 
Cor