A weighted least squares solution for space intersection of spaceborne stereo SAR data

The use of stereoscopic SAR images offers an alternative to interferometric SAR for the generation of digital elevation models (DEMs). The stereo radargrammetric method is robust and can generate DEMs of sufficient accuracy to geocode SAR images. Previous work has shown that ground coordinates with accuracy of four times the resolution cell can be obtained from ERS data without using any ground control points (GCPs), where the high accuracy of the orbit and satellite position of the order of metres introduce insignificant errors into the intersection procedure. The orbit data for RADARSAT is not as accurate as that for ERS, and the perpendicular relationship between the resultant velocity vector and the resultant range vector is uncertain in terms of image geometry. Hence, it is necessary to refine the method to allow for possible errors. This paper introduces a weighted space intersection algorithm based on an analysis of the predicted errors. A radargrammetric error model for observation errors is also formulated to predict the accuracy of the algorithm. The revised method can be used without any GCPs, but this can lead to systematic errors due to less accurate orbit data, and it has been found that the use of two GCPs provides a reasonable solution. The method is insensitive to the spatial distribution of GCPs, which is often critical in traditional methods. The error statistics of the results generated from 32 independent check points, distributed through the entire SAR image, approach the predicted errors and give positional accuracy of 38 m in three dimensions.