GEOSCAN Search Results: Fastlink


TitleSubpixel-matching based on phase correlation applied to stereo-radargrammetry
AuthorZakharov, I; Toutin, T
Source32nd Canadian Symposium on Remote Sensing and 14th Congress of the Association québecoise de télédétection: monitoring a changing world, program/32e Symposium canadien sur la télédétection et 14e Congrès de l'Association québecoise de télédétection : suivi d'un monde en mutation; 2011 p. 58 Open Access logo Open Access
LinksOnline - En ligne (complete volume - volume complet, 7z, 3.96 MB)
Alt SeriesEarth Sciences Sector, Contribution Series 20110140
Meeting32nd Canadian Symposium on Remote Sensing and 14th Congress of the Association québecoise de télédétection; Sherbrooke, QC; CA; June 13-16, 2011
Mediapaper; on-line; digital
RelatedThis publication is related to Subpixel image matching based on Fourier phase correlation for Radarsat-2 stereo-radargrammetry
File formatpdf
Subjectsgeophysics; Science and Technology; remote sensing; radar methods; Methodology
ProgramTopographic Mapping
Released2011 06 01
AbstractThe purpose of the stereo-radargrammetric process consists in the extraction of spatial information from stereo synthetic aperture radar (SAR) images. The accuracy of this process depends on the values of parallax, which is the subject of the stereo pair configuration, and the accuracy in determining parallax. The goal of this work is to evaluate the effect of subpixel image matching in determining parallax for stereo-radargrammetry. To determine parallax a subpixel matching algorithm based on the phase correlation method was developed. This method calculates image cross-correlation in the Fourier domain. Upsampling a discrete Fourier transform achieves subpixel accuracy in cross-correlation peak localization without increased computational cost. A hierarchical multi-resolution approach allows the usage of a windowing algorithm to determine parallax for every image pixel. The developed algorithm was applied to Radarsat-2 data of Fine Quad (FQ) mode (5.2 x 7.6m resolution). First, we simulated a set of SAR images corresponding to various Radarsat-2 beams between FQ5 (25°) and FQ18 (38°). The experimental results with simulated images show that a decrease in intersection angle leads to an increase in matching accuracy up to 1/20 of a pixel. Second, the algorithm was applied for matching the real stereo pair FQ5 and FQ18. To validate the matching results a digital surface model was extracted and compared with accurate lidar data. The analysis of the mutual dependence of matching accuracy and stereo pair configurations shows that the application of sub-pixel matching allows us to make the radargrammetric process flexible in the choice of stereo pairs.

Date modified: