Fully automatic geo-coding of CARABAS-II VHF SAR images

Abstract

Fully automatic geo-coding algorithms for CARABAS-II VHF SAR images have been developed. By using recorded flight parameters and a coarse scale (50/spl times/50 m) digital elevation model (DEM), complex CARABAS-II slant range images were gee-coded to a local orthogonal map projection and vice versa. The aims of this paper are to describe the algorithms and to evaluate the geometric accuracy obtained. The SAR image processing was performed with a global back projection technique in order to minimise any geometrical distortions. The analysed images were processed by focusing to a nominal reference height close to an average altitude over the area of interest. With the assumption that the actual flight path is close to a straight line, the errors introduced by the approximate focusing can be negligible during normal flight conditions. The evaluation of the geo-coding accuracy was performed by measuring the displacement between curved line segments extracted from a map and corresponding visually detected segments in the CARABAS images. To avoid bias in measurements for images within single flight paths, two sets of images from different flight paths were used for the evaluation. The result shows that the image processing and the geo-coding algorithms have been successfully implemented. The mean displacements for the investigated forest stand borders, lake borders, and main stream border were found to be 3.4, 2.6, and 9.3 m respectively. This implies that the recorded flight parameters are reliable and that the system performance is stable. Hence, the algorithms can be used in both research studies and real world applications.