Geometry Processing For Remote Sensing Image Based On Expanded RPC Model

Earth observation by satellites is the important method of acquisitioning the spatial information on terrestrial space, and is producing important result on national economic construction and national defense construction. Facing the more and more multi-source Remote Sensing Image, how to develop the generalized sensor model to proces the multi-source remote sensing image unified in geometry is an urgent task.This paper focuses on the theory and method of geometry processing of remote sensing image based on expanded RPC mode. Its idea can be expressed as the following:When the RPC model is solved using the terrain-independent solution, the question is whether the over-parameterization of the RPC has led to the design matrix becoming ill conditioned and the normal matrix being singular in a row. In order to improve the condition number of the normal matrix, the ridge estimate in which a small multiplication of the identity matrix is added was applied. But the ridge estimate raises two problems: the first is that the ridge estimate changes the isometric relation in the normal matrix leading to the estimated result being a biased estimator. The second is that the determination of the ridge parameter is difficult and can be selected over a wide range. So based on the bispectrum analysis, an unbiased RPC estimator is presented which can improve the status of the normal matrix and does not change the isometric relation of the normal matrix. Also based on the finite element interpolate, an expanded RPC model is presented, the expanded RPC model is suited not only for push-broom imagery but also for SAR imagery.Then derivative the error equation of space resection, space intersection and the block adjustment based on the expanded RPC model, using these in the rectification and block adjustment for multi-source remote sensing image. In short, it has been proved by experiments that rectification and block adjustment of satellite remote sensing imagery using the expanded RPC model is successful.Also based on the finite element interpolate, 3D spatial grid controlled re-sampling scheme is successfully developed for rectification remote sensing image using the expanded RPC model, which takes the rectification accuracy affected by the hypsography.Based on the study of the algorithms, the pilot system on geometry processing of remote sensing image based on the expanded RPC model is developed, which can beused to process the SPOT-5 image, QUICKBIRD image, IKONOS image etc push broom remote sensing image and ERS image, RADARSAT image etc SAR remote sensing image.