Abstract: Aiming at the insufficient research on modeling complex indoor and underground spaces in real-scene 3D construction,this paper proposes a 3D modeling method that integrates 3D point clouds and video imagery.Firstly,high-precision laser point cloud data inside the cave is acquired using SLAM technology,while multi-angle video imagery is collected via close-range photogrammetry.Then,the SFM algorithm is employed to generate dense matching point clouds,and the ICP algorithm is applied to achieve precise registration of heterogeneous data,constructing a 3D cave model with both structural features and texture information.Finally,the indoor model is fused with an outdoor terrain-level real-scene 3D model obtained from UAV oblique photogrammetry,forming a unified digital twin platform.The results demonstrate that this method achieves high-precision reconstruction and virtual-real integration of indoor and outdoor scenes,providing a reliable technical reference for modeling complex underground spaces.

Key words: SLAM technology, close-range photogrammetry, UAV oblique photogrammetry, video and point cloud fusion modeling, 3D fusion modeling inside and outside the cave