Precise mapping of high resolution Satellite images without ground control points

초록

As one way of removing the requirement of ground control points (GCPs) for precise mapping of high resolution satellite images (HRSIs), this paper proposes to use existing low-resolution elevation dataset. We adapted DEM matching technique, originally proposed to solve absolute orientation of aerial images from perspective sensors, for establishing absolute orientation of HRSIs from pushbroom sensors without GCPs. We used a SPOT-5 stereo pair with the resolution of 2.5m and DTED level 2 data with grid spacing of 30 m for experiments. First, we established relative orientation for the pair. We showed that conventional relative orientation methods with tiepoints were not applicable to pushbroom-type images due to the intrinsic property of pushbroom geometry. We proposed to use pseudo GCPs, the artificial control points generated using existing elevation data and initial geometric models available from the satellite metadata, for relative orientation. Then, we proposed mathematical formulation of DEM matching for pushbroom images. From the image pair with relative orientation established, tiepoints and their 3D model coordinates were automatically generated. These were matched against the DTED level 2 data and the transformation between the model space and the object space was retrieved. We showed that by applying this transformation to the image pair we can establish absolute orientation of the left and right images. The major contribution of this paper is that we applied DEM matching to pushbroom images and that we verified the applicability of the existing elevation dataset available worldwide for precisely mapping HRSIs without external GCPs.