Abstract: To address the issues faced by existing 3D modeling methods,such as limited modeling parameters,low efficiency,high labor costs,and lack of details.Firstly,by integrating oblique photography 3D models,refined laser point clouds,and geographic information semantic technology,we break through the bottleneck of data fusion under complex occlusions (vegetation,buildings),providing multi-type,full-coverage elements and reliable data sources such as textures,geometry,and point clouds for modeling.Then,the 3D Gaussian splash (3DGS) modeling method is applied,utilizing anisotropic Gaussian ellipsoids to construct scenes,combined with optimization and fast rasterization techniques.And high-quality and efficient reconstruction of full-element real-world models under large elevation differences is achieved,with model accuracy reaching the centimeter level.The research shows that the 3DGS method based on multi-source surveying and mapping geographic information fusion can effectively solve the 3D modeling challenges in complex environments with high occlusions and large elevation differences,providing a new paradigm for high-quality and efficient 3D reconstruction.

Key words: multi-source surveying and mapping geographic information data, oblique photography, laser point cloud, 3D modeling, 3D Gaussian splash