Landslides would move slowly during earthquakes, rainfall and excavation at the foot of the slope, and some landslides will be directly destabilized and damaged, but some slow-moving landslides will gradually become stable and in a dormant state. At present, the external manifestations and internal mechanisms of self-healing process are still less studied, especially the self-healing characteristics of erosive slow-speed landslides at the foot of the slope. On June 17, 2020, a debris flow broke out in Meilonggou, Danba County, Sichuan Province, blocking Xiaojinchuan, forming a chain of debris flow - dammed lake - burst flood disaster. The outburst flood eroded the foot of the slope, leading to the resurrection of the ancient landslide in Anianzhai. Through the emergency monitoring data of the Anianangzhai landslide, it found that although the ancient landslide of Anianzhai had large deformation in the initial stage, the deformation gradually stabilized in the later stage, and there was a self-healing phenomenon of strength recovery. Monitoring data show that the two sudden starts of the landslide are closely related to the increasing discharge and toe erosion, and have little to do with earthquakes and rainfall. By the calculation of monitoring data, the deformation rate of the landslide is as high as 118 mm/h when it just occurs.In the absence of external triggers, our prediction model indicated that the deformation rate of the sliding body was as low as 1 mm/h around August 21. This study analyzed the deformation process and self-healing characteristics of the Anyangzhai landslide, and established a self-healing prediction model based on field monitoring data. This method is of great significance for scientific evaluation of landslide stability and reduction of landslide control design cost.