Abstract: The velocity estimation and SAR imaging are the two significant topics for the forward-looking ground penetrating radar (FLGPR). This paper presents a fast subsurface SAR imaging algorithm with nonstationary filtering matrix for FLGPR. The target image is reconstructed from FLGPR received data after a nonstationary convolution filtering in the Fourier domain. And the velocity of the electromagnetic wave propagating under ground is estimated by using the image gray variance after synthetic aperture imaging with various velocities. The effectiveness of the approach is demonstrated with simulated data and experimental data set.