在斜视合成孔径雷达（Synthetic Aperture Radar, SAR）成像的情况中，由于目标的运动或者平台运动误差，导致成像结果方位向散焦，需要通过自聚焦处理改善图像质量。但由于斜视成像时点目标扩展函数二维不正交，采用传统自聚焦算法难以提取出强点目标较完整的方位向能量进行自聚焦处理，聚焦效果不佳。本文提出了一种基于BP（Back Projection）算法成像点目标扩展函数（Point Spread Function, PSF）二维不正交的自聚焦方法，利用后向投影算法对晃动的舰船目标进行成像，通过坐标旋转解决方位向倾斜导致PGA性能不佳的问题，得到聚焦良好的图像。仿真结果验证了本文方法的有效性。

In the case of squint SAR imaging, the movement of target and the error of platform movement can generate the echo phase error, resulting in azimuth defocusing. Phase Gradient Autofocus (PGA) processing is need to improve image quality. However, the Point Spread Function is not orthogonal in two dimensions during the squint imaging. Using the traditional autofocus algorithm is hard to extract the complete azimuth energy of the strong point target for autofocus processing, leading to poor focusing. This paper proposes a 2D Non-Orthogonal Point Spread Function (PSF) autofocus algorithms for squint SAR based on BP imaging algorithm. Using BP algorithm to image the shaking ship target. Using coordinate rotation to solve the problem of poor PGA performance caused by azimuth tilt, finally we get a well-focused image. The simulation results verify the effectiveness of this method.