<p>As a geological disaster that causes great damage to the natural and social environment, the landslide susceptibility evaluation is very important for disaster prevention and mitigation. This paper takes Heiniugou Village in Pengyang County, Ningxia, China as the research area. Based on the three-dimensional real scene model constructed by unmanned aerial vehicle (UAV) tilt photogrammetry technology, the loess landslide in the study area was identified and recorded. On this basis, the landslide susceptibility evaluation and zoning are carried out by using the subjective and objective combination weighting method. The results show that the three-dimensional real scene model based on UAV oblique photogrammetry can accurately and quickly identify regional loess landslides from multiple dimensions and perspectives. A total of 21 landslides in the study area are identified, and their spatial distribution, scale, Visible Light Band Differential Vegetation Index (VDVI), slope aspect and slope and land use characteristic parameters are analyzed, which can make up for the shortcomings of current two-dimensional plane remote sensing images. Based on the combined weighting model, the proportion of landslides in the extremely high susceptibility area of the study area is 33.33%; the landslide in the high-prone area accounts for 52.39%; the landslide in the prone area accounts for 4.76%; landslides in low-prone areas accounted for 4.76%; the landslides proportion in extremely low-prone areas is 4.76%. The receiver operating characteristic curve was used to complete the test of landslide susceptibility evaluation results based on the combined weighting model (area under curve = 0.736). It can be seen that the model has high evaluation accuracy. The three-dimensional modeling technology based on UAV tilt photogrammetry provides a new idea for landslide susceptibility evaluation, and provides demonstration and reference for landslide disaster prevention and mitigation in other loess areas.</p>