A COMPARISON OF ACCURACIES OF THE RPC MODELS: HOMO- AND HETERO-TYPE STEREO PAIRS OF GEOEYE AND WORLDVIEW IMAGES

Abstract. We investigated the accuracy in three dimensional geo-positioning derived by two homo-type stereo pairs and four hetero-type stereo pairs of high resolution satellite images using the vendor-provided rational polynomial coefficients (RPC) in this research. The results of 3D geo-positioning from six different stereo combinations were assessed with seventeen GPS points which were commonly well distributed in the scenes. Recently, satellite image vendors provide homo-type stereo pair images taken by the same sensor during a short time period. Stereo pair images have good geometry for achieving accurate ground coordinates. However it is difficult to acquire them at the request time because of the revisit time of the satellite and current weather conditions. Due to these reasons, a new methodology using hetero-type stereo pairs has been suggested to derive ground coordinates. High resolution satellite images include the rational function model in the form of RPC which represents the relationship between the image coordinates and object coordinates with rational polynomials. RPC makes it fast, accurate, and simple to calculate ground coordinates without any exterior orientation parameters of satellites. We constituted six different stereo pairs from four images of GeoEye-1 in-track stereo pair images and WorldView-2 in-track stereo pair images which were collected for the same region (17.5 km x 10.0 km) of the west coast in Korea. We collected GCPs by differential GPS surveying. The ground coordinates derived from six different pairs without and with some GCPs were compared to all GPS points respectively. The accuracy of ground coordinates from hetero-type stereo pairs is equivalent to the accuracy from homo-type stereo pairs. This research demonstrates that we can achieve comparatively accurate ground coordinates without GCPs using any stereo combinations of images containing proper RPCs, although we don't have in-track stereo pair images. Furthermore, some proper combinations of images with GCPs can improve the positioning accuracy.