Abstract: Faults are geologically critical to safe coal mining and hydrocarbon enrichment. Fault identification is crucial for disaster prevention and control, migration pathway characterization, and the construction of geological transparency during exploration and development. To address the limitations of traditional coherence attributes in low signal-to-noise ratio (SNR) data volumes—such as weak fault responses, strong non-structural background interference, and discontinuous fault trajectories—a fault identification method based on an improved Frangi filter and cascaded spatial constraints is proposed. This method achieves fault identification through the synergy of Hessian geometric sharpening, structure tensor soft constraints, and data-driven iterative background suppression. A 2D geological model of coal-bearing strata with dipping complex structures is first constructed, and the effectiveness of the proposed method is verified using forward modeling data. Random noise with different SNRs is then added to the model, and the robustness of the method under low SNR conditions is validated by comparing it with traditional methods, including the Roberts, Canny, and standard Frangi filtering operators. The application to fault identification in actual coalfield seismic data demonstrates that the proposed method exhibits good performance in recovering fault spatial continuity, enhancing boundary convergence, and suppressing background noise. These results provide a reliable geophysical basis for refined fault interpretation under complex structural backgrounds and offer important guidance for achieving the geological transparency of coal mines.