Multi-sensor Satellite Image Sub-pixel Registration

One way of geocoding a remote sensing image is to register it to an existing orthoimage. A multi-sensor and multi-resolution remote sensing image sub-pixel registration scheme is proposed in this paper. In this registration scheme, point feature, three image matching algorithms (ICM, PRM, ISM) and three constraints (APC, CDM, ICC) are used. They guarantee that the precision of point matching reaches sub-pixel and the distribution of feature points are self-adapted to the area terrain. Then TIN is constructed with feature points. Finally the target image is rectified based on TIN with the improved triangle filling algorithm. Experiment results demonstrate that the registration accuracy of SPOT5 panchromatic image (2.5 m/pixel) is 2-3 m and SPOT5 multi-spectral image (10 m/pixel) is 5-6 m. Since the method has good feasibility, efficiency and accuracy, it has been used in some emergencies such as disaster evaluation, paroxysmal epidemic situation and local war.