<p>Multi-phase seismic data merging faces challenges such as inconsistent observation systems, irregular spatial sampling, and uneven offset distribution. These issues often lead to migration aliasing, migration noise, and amplitude distortion that are difficult to resolve. To address these problems, this paper proposes a five-dimensional seismic data regularization method based on the radial domain. The core idea is to construct a radial sampling grid centered at the common reflection point and perform data reconstruction in the offset-azimuth domain, making the reconstruction process conform to the physical law of radial wavefield propagation. The anti-aliasing matching pursuit Fourier interpolation algorithm is adopted for five-dimensional data reconstruction. Using six three-dimensional seismic datasets acquired at different periods in the ZT Sag of the Sichuan Basin as a test case, this paper systematically compares the proposed method with two conventional methods: gap-filling regularization and observation system-guided regularization. The results show that while improving the consistency of bin attributes, the proposed method achieves improved amplitude fidelity than the compared methods. The final migration profile shows improvement in Signal-to-Noise Ratio and clearer fault imaging. The processing results improve the Signal-to-Noise Ratio of shallow and middle layers and the clarity of structural imaging, providing an improved data foundation for subsequent interpretation and inversion. The five-dimensional radial domain regularization method demonstrates potential as a physically reasonable approach for balancing data regularity and waveform fidelity in multi-phase data merging.</p>