Abstract: To solve the technical difficulties of using unmanned aerial vehicle (UAV) inspection for monitoring the stability of domestic waste landfills, the largest Xingfeng landfill in South China was taken as the research object to collect point cloud data. Four methods, namely radial basis function, nearest neighbor, triangulated irregular network, and inverse distance weighting were used to establish the digital elevation model. By analyzing the model parameters of the stack shadow map, terrain correlation, contour map, and maximum drainage length map, the suitable point cloud collection and modeling method for landfill sites were studied. The research results show that the landfill’s surface is covered with resin coatings with strong reflection directionality, which has an obvious directionality for the reflection of LiDAR rays and a strong inhibitory effect on the testable range of UAV. At an altitude of 50 meters on a flight route, a distance of no more than 60 meters between flights is required to ensure a data collection volume above 100 points/m². In areas with high film gradients, separate air routes must be arranged. Through modeling and application effect comparison, the triangulated irregular network method has the best comprehensive performance, with a terrain elevation accuracy deviation of no more than 2 cm, and can effectively handle tasks such as terrain contour drawing and water flow simulation.