International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004 
  
4.4 Image Matching and 3-D Reconstruction 
Object space reconstruction is the prime objective of 
photogrammetry. Traditionally, this process starts by matching 
points in the image space. However, matching points encounter 
many problems due to different geometric and radiometric 
differences between the images as well as repeated signals. It is 
very hard to reduce or eliminate mismatched automatically — it 
is up to the human operator to reject them. Apart from points, 
Habib and Kelley (2001) used the MIHT strategy to estimate the 
Relative Orientation Parameters (ROP) between stereo-pair of 
images using linear features. The suggested approach was 
successful in dealing with large-scale imagery over urban areas, 
which proved to be difficult when using traditional matching 
procedures. Habib et al. (2003a) extended this approach to 
allow for the reconstruction of corresponding 3-D linear 
features, Figure 6. Such an approach can be expanded to allow 
for surface reconstruction, ortho-photo generations, and object 
recognition applications. 
  
  
(c) 
Figure 6: Left (a) and right (b) images containing matched 
linear features and the reconstructed 3-D linear 
feature (c) 
4.5 Photogrammetric and Medical 
Registration 
Image-to-Image 
With the enormous increase in earth observing satellites, there 
has been an urgent need for establishing automatic and accurate 
registration techniques of multi-source imagery with varying 
geometric and radiometric properties. Traditional image 
registration techniques require distinct points, which have to be 
identified in the imagery. However, identifying conjugate points 
in imagery with varying geometric and radiometric resolutions 
is difficult. Habib and Al-Ruzouq (2004) used linear features for 
the co-registration of scenes captured by space-borne linear 
array scanners. The MIHT strategy has been implemented to 
automatically establish the correspondence between conjugate 
linear features as well as estimate the parameters relating the 
involved scenes. Figure 7 shows straight-line fcatures digitized 
in SPOT and IKONOS scenes as well as a mosaic scene 
generated after establishing the registration. 
    
(a) (b) (c) 
Figure 7. SPOT (a) and IKONOS (b) scenes with digitized 
linear features, which are used to generate a 
composite mosaic (c) 
   
614 
Similarly, in medical images, it is often hard to find conjugate 
points, Figure 8. Linear features, as seen in the same figure, can 
be used instead to facilitate the registration between the images, 
Heidi Lert Rear 
Doa 
. ;Feothils 3 
(ad pi 
$45 Sene, Image 113 
  
Figure 8. Registration of medical images 
4.6 Surface-to-Surface Registration 
With the increasing popularity of LIDAR systems, there has 
been an interest in establishing procedures for surface-to surface 
registration for change detection applications (e.g., Habib et al., | 
2001b). Habib et al. (2004) used LIDAR features as control 
information to establish the absolute orientation of 3-D 
photogrammetric models. Photogrammetric triangulation | 
incorporating tie linear features has been used to derive 3-D 
straight-line segments relative to an arbitrary coordinate system, 
Figure 9-a. On the other hand, LIDAR linear features have been 
extracted by processing the elevation data, Figure 9-b. Finally, 
conjugate photogrammetric and LIDAR features were used to 
determine the parameters relating the photogrammetric 
coordinate system to the LIDAR reference frame (i.c., solve for 
the absolute orientation paramcters). The approach proved to be 
successful in detecting discrepancies between the involved 
surfaces. Such discrepancies can be either attributed to changes 
in the object space and/or un-accounted systematic biases in the 
data acquisition systems. 
  
(a) (b) 
Figure 9. Straight-line features obtained from photogrammetric 
(a) and LIDAR (b) systems 
5. CONCLUSIONS AND RECCOMENDATION FOR 
FUTURE RESEARCH 
This paper discussed key issues related to the incorporation of 
linear features in photogrammetric applications. First, it has 
been established that among the possible types of linear 
features, straight-line segments are the most interesting ones. 
This can be attributed to the fact that they abundantly exist in 
imagery of man-made environments. Also, free-form linear | 
features can be represented with sufficient accuracy as à 
sequence of straight-line segments (poly-lines). Moreover, 
straight-line segments are valuable for the self-calibration of 
frame cameras and the recovery of the EOP for linear array 
scanners. 
We introduced a mathematical model for incorporating linear 
features in photogrammetric and medical problems (c.g. 
automatic space resection, photogrammetric triangulation, 
Internati 
camera : 
image-to 
example, 
can be u: 
integratic 
would al 
Straight | 
photogra 
linear art 
allow for 
addition, 
EOP ass 
In additi 
solving i 
especially 
areas. 
Finally, 
features 
tasks. Su 
source i 
resolutio: 
LIDAR : 
triangula! 
systems. 
Future re 
linear fe: 
approxim 
linear tra 
be invest 
linear. fe 
change d: 
Drewniol 
automatic 
Journal c 
Fórstner, 
Standard 
ACSM/A: 
Gülch, I 
automate 
triangulat 
Remote S 
Haala, N 
river patt 
Photogra 
29(B3): € 
Habib, A 
Features 
Environn 
State Uni 
Habib, ; 
Sequence 
Photogra 
Habib, A 
of Large 
Iterated 
Photogra