LiDAR, Photogrammetry or Gaussian Splatting?

When to choose which 3D technology?

Each 3D capture technology has different strengths. The right choice depends on what you want to capture, how quickly you need the data, and how much visual detail is important.

LiDAR

Sensor-based

Choose LiDAR when geometric accuracy combined with fast data acquisition and immediate feedback are important. LiDAR directly measures distances to the environment, providing spatial information that is naturally represented in 3D. This makes it particularly well suited for creating true-to-scale point clouds and models.

On a mobile device, LiDAR also provides immediate feedback during capture and can capture when your phone is offline. You can see the 3D model being created while you scan, making it easier to identify missing areas and rescan them while you are still on site.

Photogrammetry

Image-based

Choose photogrammetry when visual detail and appearance are important.

Photogrammetry uses overlapping images to reconstruct the scene. Cameras capture information that LiDAR does not directly provide, such as color, texture, and fine visual details.

Rather than treating LiDAR and photogrammetry as competing technologies, they can therefore be used together. LiDAR provides the geometric foundation and scale, while image data adds visual information and can help improve the positioning of the capture.

Wouldn’t it be best to combine these technologies then?

Definitely!

The strengths of LiDAR and photogrammetry complement each other.

LiDAR provides the geometric foundation with accurate spatial measurements, scale, and immediate 3D feedback. Camera data provides the visual information with color, texture, and surface details that LiDAR cannot capture as effectively.

When both data sources are processed together, the result can be more than simply a colored point cloud. The camera data can contribute to position tracking and drift correction, while LiDAR provides spatial constraints and a true-to-scale reference. This combination makes it possible to create detailed, colored, and dimensionally reliable point clouds directly on a mobile device.

That is the approach used by MAVO 3D today. MadVoxels 3D engine is combining LiDAR and camera data to create a detailed 3D representation that combines geometric accuracy with visual detail.

So where does Gaussian Splatting fit in?

Rather than replacing LiDAR and photogrammetry, Gaussian Splatting is best understood as a different way of representing and visualizing a captured environment.

Our current workflow combines LiDAR and camera data to create detailed, colored, and true-to-scale point clouds.

Gaussian Splatting takes a different approach. Instead of primarily reconstructing measurable geometry, it focuses on reproducing the visual appearance of a scene from different viewpoints.

How does Gaussian Splatting work?

3D Gaussian Splatting (3DGS) represents a scene using millions of small 3D Gaussians. Each Gaussian has properties such as its position, shape, orientation, color and opacity.

These Gaussians are created from captured images or videos so that, when viewed from different positions, they reproduce the appearance of the original scene. Because the representation is designed directly for rendering, the resulting environment can look remarkably detailed and realistic while remaining interactive.

Traditional photogrammetry can turn images into 3D points, dense point clouds, meshes, and textures that can be measured and processed as geometry. Gaussian Splatting does not need to create a conventional continuous surface. The Gaussians themselves are the representation used to render the scene.

This is also why Gaussian Splatting can look so impressive. The representation is optimized for visual appearance, rather than for producing a clean, editable surface.

For this reason, Gaussian Splatting should not be seen as a replacement for LiDAR or photogrammetry. Instead, it opens up another interesting opportunity of using accurate, detailed capture data as the foundation for highly immersive visualization.

For MAVO 3D, this makes Gaussian Splatting particularly interesting as a potential complement to our existing LiDAR + camera workflow. Today, we focus on creating true-to-scale, detailed point clouds. In the future, Gaussian Splatting could add a highly realistic visual layer on top of that captured environment.

Gaussian Splatting

Image-based

Choose Gaussian Splatting when visual realism and interactive exploration are more important than metrically reliable geometry.

Can you measure a Gaussian Splat?

Not reliably by default.

A Gaussian Splat does contain spatial information, because its primitives exist in 3D and have properties such as position, scale, and orientation. However, they don’t have a metrically reliable surface model.

The standard 3DGS representation is optimized primarily to reproduce the appearance of the captured images. It does not automatically provide the same kind of explicit, continuous geometry that you would obtain from a mesh or a laser-scanned point cloud.

There is active research into making Gaussian Splatting more geometry-aware and metrically accurate, including approaches that incorporate LiDAR or other geometric information.

For applications such as surveying, floor plans, dimensional documentation, BIM, or other measurement workflows, Gaussian Splatting should therefore not automatically be treated as a replacement for a metrically controlled LiDAR or photogrammetric dataset.

Does Gaussian Splatting replace photogrammetry?

No, it’s different tools for different goals.

Photogrammetry is useful to reconstruct measurable and editable geometry from images. Its outputs can include point clouds, meshes, textures that can be used for measurement and further geometric processing.

Gaussian Splatting is interesting when the goal is to create a highly realistic and interactive visual representation of a captured environment.

Research is increasingly combining Gaussian Splatting with geometric information from LiDAR, depth sensors, and other reconstruction methods.

At the moment, the main strength of Gaussian Splatting is its ability to provide an highly detailed visual experience, rather than a guaranteed metrically reliable 3D model.

Getting started

You are looking for a true-to-scale 3D scanning solution, built for as-built documentation? Download MAVO 3D on the App Store or talk to our 3D scanning experts.

Other questions?

Talk to a 3D scanning expert.

Instagram and LinkedIn