Automated HRSI georegistration using orthoimage and SRTM: Focusing KOMPSAT-2 imagery

Conventional manual georegistration has been labor-intensive and time-consuming to process globally acquired high-resolution satellite imagery (HRSI) including the Korea Multi-Purpose Satellite-2 (KOMPSAT-2). While an image-to-image matching-based method can provide an automated procedure, geometric and pixel value differences such as the apparent leaning of the terrain limits an accurate georegistration. This study utilizes the projection of orthoimages into the HRSI space using external elevation data to reduce the geometric difference, and an edge matching is applied to overcome the pixel value differences. Either rigorous or replacement sensor modeling is then carried out for georegistration with robust outlier removal. An experiment for a KOMPSAT-2 strip using Digital Orthophoto Quadrangles and the Shuttle Radar Topography Mission improved the horizontal positional accuracy from 70 to only a few meters for the test data set.

► An automatic georegistration of high-resolution satellite imagery is proposed.
► We project orthophotos into KOMPSAT-2 image to reduce the geometric difference.
► We use an edge matching technique to overcome the spectral difference.
► A robust outlier removal technique is applied to georegistration.
► Accuracy significantly improved from 70 to only a few meters for the test data set.