Research on the Application of 3D Gaussian Splatting in Railway Three-Dimensional Visualization Technology
To improve the informatization level of railway engineering construction and serve the Smart Railway 2.0 initiative,the Shenzhen to Shantou railway section 2 informatization project,which relies on the railway engineering management platform,has researched 3D Gaussian splatting-based railway 3D visualization technology.Firstly,a technical route for railway 3D visualization based on 3D Gaussian splatting was established,utilizing a three-layer structure consisting of the data layer,regional 3D Gaussian splatting field,and rendering layer.This approach involves drone data collection,3D Gaussian splatting machine learning,SuperSplat shearing and assembling,and finally,implementing project applications using Three.js-Gaussian Splats3D.Secondly,the focus was on analyzing the regional training mechanism of 3D Gaussian splatting for long and large scenarios,with a 10m buffer zone for regional training of long and large railway projects;SFM point cloud processing;creation of 3D Gaussian ellipsoid sets;machine learning training based on PyTorch;and 3D Gaussian rasterization processing mechanism.Finally,a pilot application of 3D Gaussian splatting was conducted on the Shenzhen to Shantou railway project,verifying the feasibility of this technical route and analyzing its rendering advantages compared with traditional oblique photogrammetry 3D reconstruction.This provides a new technical route for railway 3D reconstruction.
3D Gaussian SplattingregionalizationSFMgaussian ellipsoid setrasterization
万方数据
