I
I
663
In: Wagner W., Szekely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B
demonstrates the detail of the accuracy assessment. From and resulted in 3.30 meters, -1.78 meters, and -3.60 meters
this table, we can find that only one of twenty five points deviation in x, y, and z dimension,
has deviation from the manually measured correspondences
Table 2. Accuracy evaluation of DSM
No
Tie points Extracted by algorithm
Tie points extracted manually
Deviation
Column
Row
Column
Row
Column
Row
Column
Row
Ax(m)
Ay(m)
Az(m)
1
2594
551
2654
546
2594
551
2654
546
0
0
0
2
4664
480
4701
571
4664
480
4701
571
0
0
0
3
7904
690
7940
791
7904
690
7940
791
0
0
0
4
1381
1938
1450
1903
1381
1938
1447
1903
0
0
0
5
4478
1565
4518
1648
4478
1565
4518
1648
0
0
0
6
6842
1549
6880
1645
6842
1549
6880
1645
0
0
0
7
1245
2376
1309
2347
1245
2376
1309
2347
0
0
0
8
3064
2927
3109
2988
3064
2927
3109
2988
0
0
0
9
6196
2412
6237
2488
6196
2412
6237
2488
0
0
0
10
177
3521
246
3479
177
3521
246
3479
0
0
0
11
3083
3310
3127
3371
3083
3310
3127
3371
0
0
0
12
5837
3903
5881
3964
5837
3903
5881
3964
0
0
0
13
929
4879
995
4839
929
4879
995
4839
0
0
0
14
2532
4965
2597
4941
2532
4965
2597
4941
0
0
0
15
5321
4839
5369
4885
5321
4839
5369
4885
0
0
0
16
7850
4150
7891
4232
7850
4150
7891
4232
0
0
0
17
2153
5575
2225
5515
2153
5575
2225
5515
0
0
0
18
4826
5352
4877
5380
4826
5352
4877
5380
0
0
0
19
7815
5005
7859
5062
7815
5005
7859
5062
0
0
0
20
5114
298
5151
393
5114
298
5150
395
0
0
0
21
2982
1621
3036
1640
2982
1621
3036
1640
3.30
-1.78
-3.6
22
6944
1740
6990
1838
6944
1740
6982
1833
0
0
0
23
5070
2186
5111
2265
5070
2186
5111
2265
0
0
0
24
2246
3935
2296
3974
2246
3935
2296
3974
0
0
0
25
4368
4772
4426
4776
4368
4772
4426
4776
0
0
0
4. CONCLUSIONS
Based on the experiment and accuracy evaluation, we draw
following conclusions.
Without the help of eppipolar condition, the proposed
algorithm can process homogeneous area with the assistant
of control netwok and.
With the help of control network and the pyramid structure,
the proposed algorithm can process the large relief
variation of the satellite images.
The proposed algorithm can successfully generate DSM
from the complicated area which includes both urban area
and high mountain area. And the accuracy of the DSM is
encouraging.
Therefore, we believe this technology has the potential to
be used for 3D reconstruction in case of no camera
geometry can be used to constrain the image matching to
deal with the ambiguity in the homogeneous area.
This is an initial attempt of this technology. We plan to
conduct extensive test to this algorithm in the near future.
ACKNOWLEDGMENTS
This research is sponsored by the Natural Sciences and
Engineering Research Council of Canada through the
Discovery Grant (NSERC Discovery Grant) awarded to Dr.
Yun Zhang, the second author of the paper. We would like
to acknowledge that Professor Clive Fraser, Department of
Geomatics Engineering, University of Melbourne, Australia,
provided us with satellite images and ground control points
for the tests.
REFERENCES:
Omitted.