Adjustability and error propagation for true replacement sensor models

Abstract

True replacement sensor models, or TRSM, are those based on construction of dense object–image grids using the rigorous physical sensor model. Photogrammetric exploitation of image sensing applies the TRSMs since, like the physical models, they possess the same three important characteristics: (1) Accurate ground-to-image function; (2) Rigorous error propagation that is essentially of the same accuracy as the physical model; and, (3) Adjustability, or the ability to upgrade the TRSM parameters when additional control information becomes available after replacing the physical model. The ground-to-image functions are commonly achieved via fitting rational polynomial coefficients, RPC, to the dense grids which encompass the entire ground volume covered by the image under consideration. A novel approach for rigorous error propagation, without using added parameters, has been developed at Purdue University. The approach resolves the problem of rank deficiency of the covariance matrix associated with RPC by theeigen-values and eigen-vectors approach. This paper summarizes the new approach and presents further development to address the adjustability characteristic. Results from its application to imagery from an aerial frame camera, an airborne pushbroom sensor, and a spaceborne linear array sensor, are presented for both simulated as well as real image data. The results show essentially negligible differences when compared to those from the rigorous physical sensor models.