Abstract: This paper proposes a compensated back projection (BP) algorithm based on range-Doppler (R-D) domain grid mapping to address clutter broadening and target defocusing issues caused by high-frequency nonlinear platform maneuvers in the airborne multi-channel wide-area surveillance ground moving target detection (WAS-GMTI) system. The method first estimates the Doppler center using inertial navigation data and deblurs it, constructing a misaligned R-D two-dimensional spatial mapping grid. Second, it establishes a spatial geometric inverse mapping model, performing high-precision motion compensation and range cell movement correction pulse-by-pulse in the time domain. Finally, it eliminates the non-stationary phase error caused by platform maneuvers. Compared with traditional frequency domain preprocessing architectures, this algorithm overcomes the spatial mismatch caused by platform variable speed motion from a physical level, providing a high-fidelity data foundation for subsequent STAP processing. Simulation experiments show that under normal speed jitter intensity, traditional methods suffer from severe target misses, while this method can still maintain a moving target detection rate of over 90%, and the average deviation of azimuth estimation is controlled within one-tenth of the beamwidth, significantly improving the robustness and accuracy of parameter estimation.