[Purposes] To address the problem that point cloud data acquired during UAV LiDAR flights are prone to void areas, resulting in local missing regions in the generated digital surface model (DSM) and digital elevation model (DEM).[Methods] A void-filling method for UAV LiDAR point clouds was proposed, which operates directly in the 3D point cloud space while incorporating neighborhood slope constraints. First, void boundaries were identified based on point density derived from neighborhood statistical analysis. Then, a surface fitting method constrained by neighborhood slope information was employed, and an adaptive computation method for slope-constrained weight coefficients was proposed to achieve LiDAR point cloud void filling under complex terrain conditions. Taking the Jiahe Quarry in Hengshan County, Hunan Province as the experimental area, five representative methods were applied for comparative analysis of the point cloud void filling results.[Findings] The proposed method achieved the best performance across all three scenarios. In particular, in areas with complex terrain, the filling accuracy was improved compared to all five contrast methods.[Conclusions] The proposed method performs void interpolation directly in the 3D point cloud space, avoiding error accumulation during the conversion of gridded data structures. By introducing neighborhood slope information constraints into the interpolation process, the method fully accounts for real surface micro-topographic features, enabling more realistic and reasonable terrain reconstruction in void regions. The proposed adaptive computation method for slope-constrained weight coefficients can automatically adjust weight coefficients according to neighborhood slope variations, eliminating the need for manual parameter tuning and significantly improving the computational efficiency of void filling.