In this paper, a landslide safety evaluation model is established based on the cloud model and the improved D-S evidence theory. With the basis of multi-source information fusion theory, combined with the safety monitoring data and landslide deformation and damage characteristics, a perfect landslide safety evaluation system is constructed, the cloud model is applied to generate the affiliation degree of each evaluation index, and the improved D-S evidence theory with the introduction of weight coefficients is used for the data fusion analysis to realize a comprehensive evaluation of landslide safety. The safety evaluation analysis of Zhoujia landslide in the Yalong River Kala Hydropower Station is studied as an example, and the safety status of the Zhoujia landslide is determined according to the principle of maximum affiliation. The results show that the whole area is in a basically stable state, while area B1 is in a unstable state, which is basically consistent with the actual status.