With the development of space-borne SAR and wide application of SAR image, the accurate positioning in geometric rectification of SAR image becomes more and more important. It corrects the image by using orbit parameters and RD location model. The orbit parameter accuracy affects the accuracy of geo-coding directly. In this paper, a new orbit fitting algorithm based on householder transformation is presented. A comparison which uses ENVISAT images as the experimental data is made between householder transformation algorithm and the traditional least squares polynomial fitting, the result shows that the new algorithm in this paper not only avoids the ill condition of traditional least squares polynomial fitting, but also have a higher positioning accuracy.