ISPRS Workshop on Service and Application of Spatial Data Infrastructure, XXXVI(4/W6), Oct.14-16, Hangzhou, China
Remote Sensing Imagery, from the formerly descript
experiments and other experiments not presented here, some
conclusions can be drawn. The image matching becomes much
harder on pushbroom images. The new algorithm can select a
unique point from several candidates point with both global
constraint and local constraint.
There still exist a number of ways to improve our algorithm:
For example the precision of the final estimation of the epipolar
geometric depends tightly on those of the 2D matched points.
To have a better estimation of the epipolar geometric, we should
increase the accuracy of the matched points and make the good
distribution of the 2D matched points. For example, extracting
certain SUSAN points in every block in the left image and
deselecting some matched point where many points in one
block are good method.
REFERENCES
Jiang W.S., Zhang Jianqing, Zhang Zuxun, 2002. Simulation of
Three line CCD Sattellite Images From Given Orthoimage and
DEM. Geomatics and Information Science Of Wuhan
University, 27(4).
Tian Q. and Huhns M., 1986. Algorithms for subpixel
registration. Computer Vision, Graphics, and Image
Processing, vol. 35, pp. 220--233.
Wang X.Y., Wang X.L., 1999. Improve and implement a
heuristic search technique—hill-climbing. Journal of Shaanxi
Normal University (Natural Science Edition), 27(1).
Zhang Guo, 2005. Rectification for High Resolution Remote
Sensing Image Under Lack of Ground Control Points. PHD
Dissertation of Wuhan university.
Zhang Z.X., 2000. Digital photogrametry. Wuhan University
Press, China.
http://www.photogrammetry.ethz.ch/general/
persons/maria/matching.pdf.
http://www.cfar.umd.edu/~kanungo/cmsc828c/kim/slide.ppt.
http://www.fmrib.ox.ac.Uk/~steve/susan/http.7/www.inria.ff/rrrt/
rr-2273.html.
http://www.unikarlsruhe.de/Uni/RZ/Forschung/Numerik/vecfe
m/guide/xvem/help/Tutorial.FEM.html.
http://www.inrialpes.ff/movi/people/Triggs/p/Triggs-iccv01-
subpix.pdf.
