In: Wagner W., Székely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B
3.1 Experimental evaluation
We tested our algorithm using the simulated images
emphasizing the following: 1.Temporal changes simulated by
adding and removing structures and lines, 2. Multi- sensor data
simulated by different spatial resolution (rotation and scaling).
The evaluation criterion is the repeatability score. Thus, the test
sequences comprise images of real textured and structured
scenes. Figure 2 shows that TPE with reasonable accuracy rate
of >0.9 maintained for temporal change rate of < 40% of 26
simulated scenario of spatial variances including feature erasure
and displacement. For cardinal changes (> 40%) the RMS
threshold in stage 2 (topological map-matching) fails to identify
the correct CPs pair.
Figure 2. Temporal change versus registration accuracy, blue
points are different simulation of temporal changes, black
hatched line is trend line, and red cursor is RMS threshold of
topological map-matching stage
An artificial scaling and rotation of simulated image evaluates
matching accuracy for multi-sensor dataset. Table 1 summarizes
the error in displacement, where TPE error represents Test Point
error for 10% of all CPs pairs. The "Original" corresponds to
original spatial resolution (0.1m) and orientation (0°) of
simulated image, "Siml" corresponds to rotation of 100°,
"Sim 11" corresponds to rotation of 100° and scaling of X2,
"Sim2" corresponds to rotation of 290°, "Sim2_2" corresponds
to rotation of 290° and scaling of X2.5.
Simulated dataset
Original
Siml
Siml_l
Sim2
Sim2_2
TPE
0.00001
0.00032
0.0056
0.00038
0.0063
Table 1. Error (in m) for simulated multi-sensor dataset when
original image resolution is 0.1 m and orientation is 0°
3.2 Case Study
In this study we operated three sensors emphasizing multi-
sensor registration in two selected time domains where multi
temporal changes were occurred. The selected sensors
documented in table 2. Images of three sensors (table 2) was an
area of 1.5X1.1 km in mid (33°30' / 34°42') Israel.
Table 3 summarizes the error in displacement of three images
(Panchromatic scanner 1, and two Panchromatic scanner 2
images) to Ikonos image from 2008 where TPE error represents
Test Point error for 10% of all CPs pairs.
Simulated dataset
TPE
Panchromatic scanner 1
0.001
Panchromatic scanner 2
0.0029
Panchromatic scanner 2
0.0035
Table 3. Error (in m) for simulated multi-sensor dataset when
original images resolution is 1, 0.25, 0.12 m
4. CONCLUSION
We propose an AIRTop algorithm as method for the solution of
the core problem of multi-sensor and multi-temporal images
registration, based on combination between SURF (Speeded Up
Robust Features) method and weight-based topological map
matching algorithm (tMM). The main focus of our algorithm is
on scale and image rotation invariant the quality of the scanning
system. Both simulated experimental and real-world case study
results shows high accuracy of registration process.
5. REFERENCES
1. Berg, M., Kreveld, M., 1997. Trekking in the Alps without
Freezing or Getting Tired. Algorithmica, 18, 306-323.
2. Brown, L. G., 1992. A survey of image registration
techniques, ACM Computing Surveys. 24(4), 325-376.
3. Brown, H., Lowe, D., Invariant features from interest point
groups, in BMVC, 2002.
4. Canny, J., 1986. A computational approach to edge detection.
IEEE Transactions on Pattern Analysis and Machine
Intelligence, 8(6), 679-698.
5. Duda, R.O., Hart, P.E., 1975. Use of the Hough transform to
detect lines and curves in pictures, Association for Computing
Machinery, 15, 11-15.
6. Greenfeld, J.S., 2002. Matching GPS observation to location
on a digital map. In: Preceedings of 81st Annual Meeting of the
Transportation Research Board, Washington, DC.
7Jensen, J. R., 2004. Introductory digital image processing,
3rd Ed., Upper Saddle River, NJ: Prentice Hall.
8.Lindeberg, T. 2004. Feature detection with automatic scale
selection. "International Journal of Computer Vision", 30(2),
79-116.
Sensor
Type
Detector
Spatial
Resolution
Radiometric
resolution
Date
Ikonos
Spacebome
Pushbroom
1 m
11 bit
06.2008
Panchromatic
scanner 1
Airborne
Pushbroom
0.25 m
12 bit
02.2009
Panchromatic
scanner 2
Airborne
Whiskbroom
0.12m
8 bit
06.2009
Panchromatic
scanner 2
Airborne
Whiskbroom
0.12m
8 bit
07.2009
9. Moigne, J. L., W. J. Campbell, and R. F. Cromp, 2002. "An
automated parallel image registration technique based on the
correlation of wavelet features". IEEE Transaction on
Geoscience and Remote Sensing, 40(8), 1849-1864.
10. Nelson, S.J., Day, M.R., Buffone, P., Wald, L.L., Budinger,
T.F., Hawkins, R., Dillon, W., Huhn, S., Prados, M., Chang, S.,
Vigneron, D.B., 1997. Alignment of volume mri and high
resolution f-18 flurodeoxyglu-cose pet images for evaluation of
Table 2. Selected sensors for case study