Abstract: It is difficult for SAR to fly as the nominal line because of atmosphere turbulence etc. As a result, the defocus and geometric distortion appears in SAR image, which mainly happens on the azimuth. So, the elimination of azimuth geometric distortion of SAR image plays an important role in SAR image application. This paper analyzes the influence of trajectory deviations to the azimuth geometric distortion of SAR image, and builds the model of the familiar error in line of sight. The accelerating model and sine model are build according to the model of range. The paper discusses the sine model in more detail, and deduces its orthogonal polynomial spread, getting its prediction expression and rectifying phase of azimuth geometric distortion. In the end, the results of the simulation and real SAR data processing testify the algorithm’s validity.