The ultra-deep strike-slip fault are one of the significant oil and gas enrichment belts in the Tarim Basin. The formation of the carbonate rock dissolution cave-type reservoirs in this area and the process of oil and gas accumulation are closely related to the corrosive and transformative actions of the strike-slip faults. Existing data indicate that strike-slip faults exhibit uneven energy and chaotic seismic reflection characteristics on profiles. Accurately identifying the strike-slip faults within the ultra-deep dense limestone is crucial for enhancing the productivity of oil fields. Conventional fracture attributes work well for identifying large faults and are widely used in faulted basins. However, in the Tarim Basin, conventional methods such as coherence and curvature have lower identification accuracy for faults of different scales. This paper proposes a combined method to improve the identification accuracy of strike-slip faults. Initially, seismic data from the F area in the northern Tarim Basin is selected, and the original data undergoes structure-guided smoothing filtering to enhance the signal-to-noise ratio of the seismic data. The denoised data is then subjected to broad-band processing to suppress low-frequency noise and expand high frequencies, effectively improving the clarity of fault points. Based on the interpretive processing, AI fault attribute recognition based on expert labels was carried out on the data, effectively improving the identification accuracy of strike-slip faults at different scales in the F region of Tabei, providing strong support for regional oil and gas exploration.