• The only information introduced to the matching step
is the image size, nominal accuracy, and the coordi-
nates of end-points of true line-segments in both image
and object spaces. Therefore, there is no need for any
initial information such as
IOP
/
EOP
and
RPC
s.
• Using a
HQPS
strategy in image space as the first step
makes the proposed
SLIM
be convergent very fast even
in a very limited number of iterations. This property
decreases the computational time significantly and
makes it practical and general. In other words, one of
the key reasons of the high reliability of the
SLIM
results
is to introduce the
HQPS
-phase. Based on this proce-
dure, only high-qualified well distributed patterns
which are more likely to have any correspondence in
object space are selected. These patterns are selected
based on inherent characteristics of the used data. So,
there is no need to any initial approximations in this
method.
• Additionally, the first and second thresholds in the
proposed screening-procedure (matching-phase) are
only used to decrease the search space as well as the
computational time. Therefore, they do not change the
final results directly.
• The third threshold in screening-procedure (matching-
phase) makes the proposed method be reliable and
accurate even in areas with repetitive patterns. So, it is
the only key threshold should be tuned based on the
nominal accuracy of the high resolution images. This
threshold also speeds up the procedure implicitly.
• Introducing the third step (final-phase) improves the
ability of the
SLIM
to find approximately all possible
inliers. So, it increases the capacity-factors more than
80 percent.
• Combination of all steps in the
SLIM
makes the algo-
rithm more reliable as well as accurate and time benefi-
cial than other classic methods (such as
RANSAC
).
• The above results show the ability of the proposed
method to find the corresponding lines for high resolu-
tion images accurately and automatically. In spite of
the above mentioned advantages of the
SLIM
, the only
weakness of this model could be its dependency to the
threshold of
T
3
.
Conclusions
In this paper, the issue of automatic high resolution image to
map matching based on linear features through introducing
new concepts of
MGL
s and
MGP
s, and a novel
SLIM
method has
been investigated. Based on the proposed strategy, only very
limited numbers of high-quality-patterns in first space (e.g.,
image space) are weighted and selected using a new
HQPS
strategy. These patterns are introduced to the matching-phase
to determine the correspondence. This step uses a screening-
procedure to prune improper patterns in second space (e.g.,
object space). The correspondence of crossing-lines is also
determined using these proposed screening procedures in the
same manner. The matched-lines are introduced to the final-
phase to find the best unique results. Based on the results, the
advantages of the proposed
SLIM
are: (a) no need for initial ap-
proximations or information; (b) simplicity of the implemen-
tation process; (c) very efficient computational time by divid-
ing the whole process to smaller parts as feasible without any
reduction on the accessible accuracy; (d) no need for human
operations; (e) generality of the proposed model for use in dif-
ferent sensors without any need to change the basic concepts;
(f) use of the most inherent conceivable information of the
extracted features; and (g) maximum reachable accuracy and
reliability. Compared to the
RANSAC
, the proposed
SLIM
has
superiority in finding approximately all conjugate lines (up
to 80 percent of inliers) with 100 percent of precision with a
very low computational time. The only weakness of proposed
SLIM
could be its dependency to the threshold of
T
3
. Based on
the results, the proposed method may improve the potential
of matching in terms of automation, accuracy, reliability, and
reduction of computational complexity and time.
Acknowledgments
We are grateful to the NCC for making available the vector
map over the Uromieh City.
References
Akav, A., G.H. Zalmanson, and Y. Doytsher, Y., 2004. Linear feature
based aerial triangulation,
Proceedings of the International
Archives of the Photogrammetry
,
Remote Sensing and Spatial
Information Sciences
, 12-23 July, Istanbul, Turkey, 35(B/3):7–12.
Chen, M., and Z. Shao, 2013. Robust affine-invariant line matching
for high resolution remote sensing images,
Photogrammetric
Engineering & Remote Sensing
, 79(8):753–760.
Fischler, M.A., and R.C. Bolles, 1981. random sample consensus:
a paradigm for model fitting with applications to image
analysis and automated cartography,
Communications of the
ACM
, 24(6):381–395. Fraser, C., and H. Hanley, 2003. Bias
compensation in rational functions for Ikonos satellite imagery,
Photogrammetric Engineering & Remote Sensing
, 69(1):53–57.
Fraser, C.S., and T. Yamakawa, 2004. Insights into the affine model
for high-resolution satellite sensor orientation,
ISPRS Journal of
Photogrammetry & Remote Sensing
, 58:275– 288.
Habib, A., and D. Kelley, 2001. Automatic relative orientation of
large scale imagery over urban areas using Modified Iterated
Hough Transform,
ISPRS Journal of Photogrammetry and Remote
Sensing
, 56:29–41.
Habib, A., M. Morgan, E.M. Kim, and R. Cheng, 2004. Linear
features in photogrammetric activities,
Proceedings of the XX
th
ISPRS Congress
, Istanbul, Turkey, Automated Geo-Spatial Data
Production and Updating Session, 12-23 July, pp.610.
Han, Y.K., Y.G. Byun, J.W. Choi, D.Y. Han, and Y.I. Kim, 2012.
Automatic registration of high-resolution images using local
properties of features,
Photogrammetric Engineering & Remote
Sensing
, 78(3):211–221.
Heuvel, F., 2003.
Automation in Architectural Photogrammetry
, PhD
thesis, Publications on Geodesy 54, NCG, Netherlands Geodetic
Commission, Delft, 90 p.
Hu, H., Q. Zhu, Z. Du, Y. Zhang, and Y. Ding, 2015. Reliable spatial
relationship constrained feature point matching of oblique
aerial images,
Photogrammetric Engineering & Remote Sensing
,
81(1):49–58, doi: 10.14358/PERS.81.1.49.
Goshtasby, A.A., 2005.
2-D and 3-D Image Registration for Medical,
Remote Sensing, and Industrial Applications
, Wiley.
Goncalves, H., 2011. Automatic image registration through image
segmentation and SIFT,
IEEE Transactions on Geoscience
and Remote Sensing
, 49(7):2589–2600, doi: 10.1109/
TGRS.2011.2109389.
Jaw, J., and N. Perny, 2008. Line feature correspondence between
object space and image space,
Photogrammetric Engineering &
Remote Sensing
, 74(12):1521–1528.
Jaw, J., and Y. Wu, 2006. Control patches for automatic single photo
orientation,
Photogrammetric Engineering & Remote Sensing
,
72(2):151–157.
Junior, J.M., and A.M.G. Tommaselli, 2013. Exterior orientation
of CBERS-2B imagery using multi-feature control and orbital
data,
ISPRS Journal of Photogrammetry and Remote Sensing
,
79:219–225.
Kang, Z., F. Jia, and L. Zhang, 2014. A robust image matching method
based on optimized BaySAC,
Photogrammetric Engineering
& Remote Sensing,
80(11):1041–1052, doi: 10.14358/
PERS.80.11.1041.
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
May 2016
375
