With regards to the passive synthetic aperture radar system using global navigation satellite system-reflectometry signals(GNSS-R),the back projection algorithm,which is usually used for imaging,has the disadvantage of high computational complexity caused by point-by-point accurate focusing.To solve this problem,an improved back projection imaging algorithm based on partitioning data and constructing phase compensation kernel is proposed.The improved algorithm first splits the echo data matrix,then divides the longest and shortest ranges of the imaging grid points in the target area into equal intervals,and then establishes a phase compensation kernel.The phase compensation results of each grid point in the sub-matrix are obtained by range search operation,and the sub images are formed by coherent accumulation in the azimuth direction.Finally,the sub images are recursively merged to obtain the imaging results of the target area.The quantitative analysis shows that the improved algorithm can effectively reduce computa-tional complexity and save memory.The simulation results verify the effectiveness of the proposed algorithm.