Photogrammetry, LiDAR, and 3D Gaussian Splatting all produce 3D data, but they answer different questions. LiDAR is about measurement. 3DGS is about visual realism. Photogrammetry sits in between, trading hardware cost for capture discipline. Picking the right one starts with the deliverable, not the device: decide what the downstream team needs to do with the data, then work backwards to the capture method.
The three methods at a glance
| Photogrammetry | LiDAR (handheld SLAM) | 3DGS | |
|---|---|---|---|
| How it works | Geometry reconstructed from overlapping photos | Direct laser distance measurement while walking | Scene of colour-carrying splats trained from imagery |
| Geometric accuracy | Variable; needs ground control for reliable scale | ~1 cm relative, ~3 cm absolute with RTK (Lixel K2) | Not measurement-grade |
| Capture speed | Slow; hundreds to thousands of photos | Fast; walk the site once | Fast; walkthrough with a capture device |
| Lighting | Needs good, even light | Works in low light and darkness | Reproduces whatever light you capture |
| Processing | Heavy CPU/GPU reconstruction | LixelStudio registration and colouring | GPU training via LCC CyberColor |
| Best for | Objects, facades, drone mapping on a budget | Scan-to-BIM, as-builts, measurable mapping | Photoreal walkthroughs, marketing, XR |
Read the table as a set of defaults, not verdicts. Every row has exceptions: photogrammetry with survey control can be excellent, and a badly walked SLAM capture can be poor, but when you are scoping a job and need a starting assumption, these are the honest ones. The sections below unpack the rows that generate the most argument.
How accurate is LiDAR vs photogrammetry?
LiDAR measures distance directly: each laser return is a real range measurement, so the resulting point cloud is inherently scaled and its accuracy is a property of the instrument, not of the scene. The handheld Lixel K2 is specified at 1 cm relative accuracy and around 3 cm RMSE absolute with its built-in RTK, and it holds that on a dark plant room as readily as on a bright facade.
Photogrammetry infers geometry by triangulating features matched across photos. Done well, with strong overlap, sharp images, and surveyed ground control points, it can reach comparable accuracy. But scale is not inherent, texture-poor surfaces (plain walls, glass, water) reconstruct badly, and every departure from ideal conditions degrades the result in ways you only discover during processing. The honest framing: LiDAR accuracy is bought with the instrument; photogrammetry accuracy is earned with discipline on every single capture.
When should you use LiDAR?
Use LiDAR whenever a number from the data will be trusted, as-built documentation, renovation measurement, Scan-to-BIM, clash checking, deformation and progress comparison. It is also the default indoors, where photogrammetry struggles with uneven light and featureless walls, and on occupied sites where you need the capture done in a short access window. A handheld SLAM walk covers a multi-storey interior in a session, works in low light, and produces registered data the same day. Even an iPhone LiDAR has a place for rough volumes: see Lixel K2 vs iPhone LiDAR for where the line sits, but once tolerances are stated in centimetres, dedicated hardware is the answer.
Where drone photogrammetry hits its limits
Drone photogrammetry earns its popularity on open, textured sites: earthworks, roofs, stockpiles. Its limits are structural. Vertical and overhanging surfaces need oblique flight planning; anything under canopy, under a soffit, or indoors is invisible to it; thin structures such as railings and masts reconstruct poorly; and in Singapore, flight authorisation constrains where and when you can fly at all. The common professional pattern is hybrid: fly the roof and open ground, walk the interiors and tight external corridors with a handheld scanner, and register the two datasets together. RTK-equipped hardware on both sides makes that registration far less painful, because each dataset already sits close to its true coordinates before alignment starts.
Gaussian splatting vs photogrammetry
Both start from imagery, but they optimise for different endpoints. Photogrammetry outputs geometry: a point cloud or textured mesh you can measure and edit. 3DGS outputs appearance: millions of oriented, semi-transparent splats that render photoreal views in a browser, including reflections, foliage, and fine detail that mesh reconstruction smooths away. Where photogrammetry needs perfect conditions to look good, 3DGS looks convincing from an ordinary walkthrough capture. The cost is that a splat scene is not a metric deliverable. Our 3DGS capture guide covers the workflow in depth.
What does each method cost?
Photogrammetry has the lowest entry cost: a camera or drone you may already own, plus software and considerable processing time. LiDAR and 3DGS need dedicated hardware, which is precisely what rental solves. Volumet's weekly dry-rental rates in Singapore (GST-exclusive): the PortalCam 3DGS camera at S$1,200, the Matterport Pro3 at S$2,000, the Lixel K1 at S$2,200, the Lixel K2 at S$2,300, the Leica BLK360 G2 at S$2,800, and the long-range Lixel L2 Pro at S$3,500. First rentals add a one-time onboarding session at S$250 (self-pickup) or S$400 (delivered). For a week of site work, the gap between methods is small next to the cost of a wrong deliverable or a repeat visit.
One capture, two outputs
You do not have to pick just one method. A single Lixel K2 walk records LiDAR and imagery together: the same capture processes into a measurable point cloud and mesh in LixelStudio, and into a photoreal 3DGS scene through LCC CyberColor. The engineers get geometry, the marketing team gets the walkthrough, and the site was only visited once. Check availability on the rental page or tell us your deliverable and we will recommend the right device.