What Is Gaussian Splatting?

In 3D visualisation, digital twin production and virtual reality (VR), there is a term we have been hearing a lot recently: Gaussian splatting. So what is this technology? Where, how and why is it used? In this article we look at Gaussian splatting and compare it with classic photogrammetry.

Gaussian splatting is a modern method for representing a scene in 3D by placing Gaussian functions (3D Gaussian distributions) on top of point clouds. Each point carries not just a position but also colour, density and spread information. This makes it possible to produce models that look much more realistic, have soft edges and render quickly.

Compared with AI-based systems such as neural radiance fields (NeRF), this method gives results faster and can run with less computing power.

How Does Gaussian Splatting Work?

Gaussian splatting works in these steps:

Image capture: photos or video taken from different angles are used.
Camera calibration: the angles from which the images were taken are determined.
Point cloud generation: a dense 3D point cloud is created from the images.
Gaussian modelling: each point is smoothed and coloured with Gaussian functions.
Real-time visualisation: GPU-based systems can render the result in real time.

Differences Between Gaussian Splatting and Photogrammetry

The comparison below shows the same area modelled with classic photogrammetry and with Gaussian splatting:

Feature	Classic photogrammetry	Gaussian splatting
Data source	Photos / drone images	Photos / drone images
3D modelling	Mesh + texture	Points + Gaussian functions
Edge sharpness	High (but with hard transitions)	Soft, natural transitions
Processing time	Medium	Low
Real-time use	Limited	Possible
GPU requirement	Medium to high	Low to medium

Where Is Gaussian Splatting Used?

Mapping and cadastral work
Construction and mine site visualisation
Architectural and urban modelling
Video games and VR applications
Cultural heritage and archaeological documentation

Advantages and Outlook

Gaussian splatting stands out in projects that need a fast workflow and real-time presentation. Its highly detailed visualisation at low computing cost gives it clear advantages over traditional mesh-based models.

In the coming years, as this method is integrated into GIS (geographic information systems) and surveying engineering projects, much faster and more intuitive systems are expected to emerge.

Conclusion

Gaussian splatting represents an important technological leap, especially in surveying engineering and visual digitisation. It is a powerful tool for professionals who want to produce high-quality 3D data with low hardware costs.

If you would like to use innovative technologies such as Gaussian splatting in your projects, you can contact us.