The RPC model has recently aroused considerable interest in the community of photogrammetry and remote sensing. The RPC is a generalized sensor model capable of achieving high approximation accuracy. Unfortunately, the computation of the parameters of RPC model is subject to the initial value of the parameter in all the literature available. In this paper, an algorithm for parameters of RPC model without initial value is presented. The algorithm was tested on SPOT-5 image. Based on numerous tests, some conclusions can be drawn. The RPC model can achieve an approximation accuracy that is extremely high for SPOT-5 pushbroom data. The results prove that the RPC model can be used as a replacement sensor model for photogrammetric restitution. When we deal with SPOT-5 data sets, the high order RPC model may be necessary in that the RPC model very much resembles the rigorous sensor model. The RPC model cases with unequal denominator can on the whole achieve better accuracy than the cases with equal denominator at check points. The RPC model cases with denominator perform better. In the establishment of the 3-D object grid for the RPC model solutions, at least two or more elevation layers are needed. For SPOT-5 imagery, when the image grid contains 21×21 point and the number of elevation layers is three, the precision and the efficiency is in balance.