levels: the object and pixel level. At the object level, most of
the obvious objects, especially the buildings and high bridges,
will be excluded from the
POA
s. At the pixel level, Dijkstra’s
shortest-path searching algorithm with a binary min-heap is
used to find the final seamlines. Two Data Sets of digital aerial
orthoimages with different ground resolution are adopted to
validate our algorithm. The results demonstrate the potential
of the proposed method. Compared with Dijkstra’s algorithm,
Chon’s algorithm and Pan’s algorithm, seamline determined
by the proposed method successfully bypass most of the obvi-
ous objects and fellow roads or rivers. Moreover, our method
has a higher efficiency and can also be integrated into the
seamlines network optimization framework easily (Pan
et al.
,
2009; Mills and McLeod, 2013; Pan
et al.
, 2014a).
Nevertheless, there is still room for improvement in the
proposed method. First, the effect of the proposed method is
determined by the result of the segmentation, and there is no
way to select the optimal parameter for segmentation auto-
matically. Second, when determining the
POA
s, our method
only used the correlation coefficient to assess the degree of
differences between objects. According to object-oriented
thinking, the attribute of the objects, such as shape, geometry,
texture, and contextual semantic information, can be used to
estimate the degree of difference between objects. These is-
sues will be addressed in the future.
Acknowledgments
The work was supported by the National Basic Research
Program of China (973 Program, No. 2014CB744201,
2012CB719901), a Foundation for the Author of National
Excellent Doctoral Dissertation of PR China (FANEDD, No.
201249), and the National Natural Science Foundation of
China (No. 91438203, 41371430, 91438112). The authors also
thank the anonymous reviews for their constructive com-
ments and suggestions.
References
Afek, Y., and A. Brand, 1998. Mosaicking of orthorectified aerial
images,
Photogrammetric Engineering & Remote Sensing
,
64(2):115–124.
Chen, M., R.A. Chowdhury, V. Ramachandran, D.L. Roche, and L.
Tong, 2007.
Priority Queues and Dijkstra’s Algorithm
, Computer
Science Department, University of Texas at Austin, 1–25 p.
Chen, Q., M. Sun, X., Hu, and Z. Zhang, 2014. Automatic seamline
network generation for urban orthophoto mosaicking with the
use of a digital surface model,
Remote Sensing
, 6(1):12334–
12359.
Cherkassky, B.V., A.V. Goldberg, and T. Radzik, 1996. Shortest paths
algorithms: Theory and experimental evaluation,
Mathematical
Programming
, 73(2):129–174.
(a)
(b)
(c)
(d)
Figure 5. Mosaicked image for Data Set 1: (a) Dijkstra’s algorithm, (b) Chon’s method, (c) Pan’s method, (d) the proposed method.
table
2.C
omparison
of
the
P
roposed
M
ethod with
D
ifferent
V
alues
of
t
for
D
ata
S
et
1
The Values of T
Number of obvious objects
passed through
Processing time(s)
0.4
none
11.669
0.5
none
11.809
0.8
1 bridge
11.965
0.9
1 bridge
11.918
128
February 2016
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
