Compared with traditional airborne phased array radar, the spatial-domain degrees-of-freedom of the airborne Transmit Subaperturing-Multiple Input Multiple Output (TS-MIMO) radar are multiplied, and the training samples required for Space-Time Adaptive Processing (STAP) are also significantly increased, so the performance degrades sharply in the actual non-uniform clutter environment, and slow moving targets cannot be detected. Different from the traditional STAP method, this paper proposes a slow moving target detection method based on two-dimensional space-time sparse super-resolution spectrum estimation. The method uses the sparse Bayesian learning algorithm to directly measure the range cell data for space-time spectrum estimation. Then, the main clutter component in the space-time super-resolution spectrum is set to zero based on the radar prior parameters, and finally the slow target can be detected in the angle-Doppler domain based onconventional constant false-alarm processing. The proposed method does not require training samples, so it can significantly improve the slow moving target detection capability of the onboard TS-MIMO radar in practical applications. Simulation experiments verify the effectiveness of the proposed method.