Automated georegistration of high-resolution satellite imagery using a RPC model with airborne lidar information

A large amount high-resolution satellite imagery (HRSI) has been available in the commercial market because of its value in creating accurate base maps for various applications. As massive amounts of HRSI are acquired globally by satellites with short revisit times, automated but accurate georegistration is still required despite advances in precise orbit tracking and estimation. Motivated by the attractive properties of airborne lidar data, such as their high resolution and accuracy, this study proposes a new automated method for refining the HRSI with rational polynomial coefficients (RPCs) using airborne lidar information. By projecting the lidar intensity return into the HRSI space, the image matching complexity is reduced to a simple, 2D case. The true challenge is in overcoming the difference between the HRSI and the lidar intensity return to allow for reliable matching. To this end, this paper proposes a new method based on simple relative edge cross correlation (RECC) with a screening method to prevent false matching. To make the approach more robust, data snooping was added for a final detection of outliers. Experiments were performed using three Kompsat-2 images and the potential of the approach was confirmed, showing sub-pixel accuracy.