What is the difference between 3DGS and photogrammetry? How to choose a 3D reconstruction route
Compare 3DGS and photogrammetry in terms of realism, geometric structure, measurement capabilities, input materials, output formats and application scenarios to help the team choose an appropriate 3D reconstruction route.
3DGS is better at quickly restoring complex appearances and real lighting, while photogrammetry is better at generating point clouds, meshes and surveying results with clear geometric structures. The two are not a simple substitute. Actual projects often need to be used in combination according to the display, measurement and delivery goals.
What scenarios are 3DGS more suitable for?
3DGS, also known as 3D Gaussian Splatting, restores real scenes through a large number of Gaussian points with position, color and transparency information. Its representation of vegetation, complex textures, small structures, and realistic lighting is generally more natural. It is suitable for space roaming, product display, cultural tourism browsing, real estate display and projects that require fast and realistic results.
What scenarios are photogrammetry more suitable for?
Photogrammetry calculates camera positions and spatial geometry from multi-view photos, and common outputs include dense point clouds, mesh models, DOMs, DSMs, and orthophotos. It is more suitable for engineering surveying, drone mapping, volume calculations, structural review and projects that require the output of standard model formats.
Should be selected backwards from the final deliverable
If the goal is a realistic and fluid view of space in a web page, 3DGS can be evaluated first. If OBJ, point cloud, DOM, DSM, 3D Tiles or engineering surveying is required, the photogrammetry or point cloud process should be prioritized. If the project requires both realistic display and engineering results, you can keep the same set of original materials and generate display results and geometric results separately in the same project.
Don’t overlook acquisition quality
No matter which algorithm is used, blur, overexposure, reflections, occlusions, insufficient overlap, and interruptions in the shooting path will affect the results. Algorithm selection is not a substitute for field collection practices.