RPC Based Image Matching for Satellite Imagery

摘要

The RFC model is a generalized sensor model mat is capable of achieving high approximation accuracy. After the RFC was verified as a replacement sensor model for push-broom image arid aerial image, several researchers assessed the accuracy of 3D positioning using the RFC. But almost all researchers studies proved that the RFC could satisfy several important requirements of the replacement sensor model to provide general tools for mapping, accurate 3D positioning, and real-time processing. Image matching for satellite imagery is not easy, especially when the geometric differences between the stereoscopic satellite images are quite large. The epipolar solution, which is popular for aerial photography, has some difficulties when applied to satellite imagery due to different geometric characteristics. This paper offers an advanced matching solution to resolve the inherent problems for the high-resolution satellites when the RFC is adopted as a basic sensor model. The proposed matching algorithm reduces the search space using object-space constraints and provides an initial position for image matching. In order to verify the applicability for high-resolution satellite imagery and check the performance of the proposed matching scheme, IKONOS stereo pairs with RFC are tested.