A quality assessment on the DEM matching-based RPC bias correction

Rational polynomial coefficients delivered by satellite image vendors facilitate the geometric processing of images. However, compensating the inherent bias of these coefficients requires in-situ ground control collection. Existing global digital elevation models are an alternative for in-situ control data collection. In this regard, the well-known DEM matching technique can be applied for aligning the image-driven DEMs with the existing global elevation models. In this study, three image-driven DEMs were generated using a commercial programming application. The commercial software, similar to the majority of other software packages, performs epipolar resampling using vendor-provided RPCs before image matching. These generated DEMs were imported to the DEM matching technique and the estimated matching parameters were utilized for correcting the bias of RPCs on the ground. We showed that epipolar resampling with vendor-provided RPCs imposes some geometric distortions to the generated DEMs and therefore, estimated that DEM matching parameters cannot be employed for RPC bias correction. In two study areas, while the geopositioning accuracy after bias correction was improved 95.6% and 93.1% using the commercial software, we achieved 97.8% and 97.4% improvement using the programming.