Ground-airborne frequency-domain electromagnettic (GAFEM) is a method that has the potential to detect underground anomalies at large depth ranges in areas with complex terrain. Currently, for the demand of field real-time rapid interpretation of GAFEM data, the apparent resistivity method cannot effectively recognize anomaly boundaries, and divergence of tipper vector imaging method has difficulties in transmitting source deployment. In view of above problems, a rapid imaging method of GAFEM based on space magnetic gradient anomaly (MGA) is proposed. This method enables the rapid recognition of underground electrical anomalies by GAFEM data without large computational resources. First, the characteristics of the magnetic field gradients in the homogeneous half-space model and the 3D anomaly model are analyzed, and the calculation method of MGA based on the background field gradient is proposed. Further, the ability to recognize underground electrical anomalies in transverse MGA, longitudinal MGA, and space MGA are analyzed by a forward model. Finally, an imaging process that can be applied to actual detection and the calculation method of background resistivity are proposed. The feasibility, the anti-noise and the multi-anomaly recognition of the imaging method are verified. The results show the imaging method based on space MGA can effectively recognize the underground electrical anomalies. The magnitude of the space MGA can determine the location and contour of the anomaly, and the real and imaginary parts can determine the electrical type of the anomaly. This study provides a new method for the interpretation of GAFEM data, which is helpful for the rapid field real-time detection of underground structures.