The geological conditions along Sichuan-Tibet Railway are complex, and frequently-occurred landslides have brought severe challenges to the railway construction. Therefore, a complete and accurate landslide perception can provide references for railway route selection and landslide risk governance. In this study, we utilized change vector analysis (CVA), principal component analysis (PCA) and independent component analysis (ICA) for change detection images (CDIs) generation, and then adopted the multi-threshold method to produce the training sample templates for landslides and non-landslides, respectively. The Markov Random Field (MRF) algorithm was further used to extract the optimal landslide objects. In particular, we tested the performance of the proposed approach using the Sentinel-2 datasets in a rapid perception of the co-seismic landslides for the Nyingchi event that occurred on 18 November 2017 and affected the railway construction. We further calculated completeness, correctness, accuracy, F1-score and Kappa coefficient, for a quantitative evaluation of landslide perception results. We found that the ICA-based change detection in MRF can extract landslides more completely and accurately. This study set up with the aim to assess the effectiveness and applicability of the proposed method in mapping landslide migration areas under complex geological conditions along the Sichuan-Tibet railway, which offers a comprehensively intelligent approach to supporting the hazard mitigation for a safe railway construction and operation.