Title :
Spatiotriangulation with multisensor HR stereo-images
Author_Institution :
Natural Resources Canada, Canada Centre for Remote Sensing, Ottawa, Ont., Canada
Abstract :
The objective of this research study was to evaluate the spatiotriangulation applied to multisensor high-resolution satellite stereo-images, which enabled the simultaneous geometric processing of a large number of stereo-pairs together to reduce the control point requirement. The spatiotriangulation is based on the three-dimensional physical models developed for multisensor images at the Canada Centre for Remote Sensing, Natural Resources Canada and on a least squares block stereo-bundle adjustment process with orbital constraints. The spatiotriangulation was applied to five along-/across-track stereo-images [panchromatic Syste`me Pour l\´Observation de la Terre 5 (SPOT-5) High-Resolution-Stereoscopy (HRS) and High-Resolution-Geometry (HRG), Ikonos, and QuickBird] acquired over Quebec, Canada. The first results of stereo/block bundle adjustment showed that the same error residuals than the input data errors (1/2 to 1-2 pixels) were obtained depending of the stereo-images, whether independently or simultaneously processed. The second and most important results were related to simultaneous stereo-bundle adjustments of the largest "master" SPOT-5 stereo-pair (either HRS or HRG) using 12 ground control points (GCPs) and the smallest "slave" stereo-pair(s) using no GCP but only stereo tie-points (TPs). Better results were normally obtained with SPOT-5-HRG (5-m resolution) as the "master" stereo-pair due to less difference in the sensors resolution than with SPOT-5 HRS. The root mean square errors, verified by independent check points (ICPs) belonging only to the "slave" stereo-pairs, were around 2 m in the three axes. However, the combined image pointing and map errors of ICPs (1-2 m) are included in these 2-m error results, and the internal accuracy of the stereo-pairs should thus be better (less than one resolution). The research study demonstrated thus the possibility to use the largest stereo-pair with a reduced number of GCPs to simultaneously adjust single or multiple stereo-pair(s) with only stereo TPs, and with no degradation in the accuracy.
Keywords :
error analysis; geophysical signal processing; image resolution; mean square error methods; sensor fusion; stereo image processing; terrain mapping; Canada Centre for Remote Sensing; Ikonos; Natural Resources Canada; Quebec; QuickBird; SPOT-5 high-resolution-geometry; SPOT-5 high-resolution-stereoscopy; control point requirement; data acquisition; error analysis; geometric processing; image resolution; independent check points; least squares block stereo-bundle adjustment; multisensor high-resolution satellite stereo-images; orbital constraints; panchromatic Systeme Pour l´Observation de la Terre 5; root mean square errors; spatiotriangulation; terrain mapping; Airborne radar; Degradation; Geometry; Image resolution; Infrared image sensors; Least squares methods; Remote sensing; Root mean square; Satellites; Strips; Error analysis; geometry; multisensor systems; remote sensing; terrain mapping;
Journal_Title :
Geoscience and Remote Sensing, IEEE Transactions on
DOI :
10.1109/TGRS.2005.861005