How it works
PhotogrammetryHow a 360 camera turns into a 3D model
If you've ever walked a job site with a 360 camera and wondered exactly what happens between "upload video" and "here's your 3D model," this is the short version of that pipeline, and why it matters for construction, engineering, and geotechnical teams.
0.1%
Typical drift
Achievable with scale targets. Survey control can bring absolute error down to 5 to 10mm.
5
Processing stages
From raw video to a textured, measurable 3D mesh, fully automated in the cloud.
1
Camera or phone
A 360 camera or a standard Android or iOS smartphone is all a field team needs.
It starts with two separate ideas
A 360 video is just a recording: a continuous, spherical view of a space as you walk through it. On its own, it's flat pixel data wrapped onto a sphere. A 3D model is something else entirely: a mathematical description of surfaces in space, built from geometry rather than pixels.
The bridge between the two is photogrammetry: the science of measuring real-world positions from photographs. The core idea is simple even though the math isn't: take the same point in space from several different angles, and triangulation can calculate exactly where that point sits in three dimensions.
From spherical frames to a point cloud
Every asset digitization run follows the same five-stage pipeline.
Feature matching
Software scans each image for identifiable points, like a crack in a wall or the edge of a pipe, and matches them across overlapping frames.
Sparse point cloud
Triangulation calculates the 3D position of each matched point, building a rough skeleton of the space.
Dense point cloud
The sparse skeleton is filled in, point by point, until the space is represented in fine detail.
Mesh generation
Points are connected into a triangulated mesh, forming continuous surfaces rather than a scatter of dots.
Texturing
Original video frames are projected back onto the mesh, so the model carries real photographic detail.
Why video instead of still photos
Traditional photogrammetry uses hundreds of individually shot, overlapping still photos. Walking a site with a 360 camera instead means every frame of a continuous video becomes usable data.
Speed of capture
A single walkthrough covers far more ground per minute than manually shooting and positioning individual photos.
Complete coverage
Capturing in every direction at once leaves a much lower chance of missing an angle or a gap in coverage.
Access to difficult spaces
A lightweight camera on a monopod reaches ceiling voids, manholes, and other awkward or unsafe spots.
The trade-off
Spherical footage spreads resolution across a wider field of view, and compression can lose fine detail, so capture quality still matters.
Origin story
Built on planetary science
The core of Sensori's photogrammetry engine did not start in construction. Co-founder Roddy O'Hara developed the underlying algorithms during his PhD research, where they were used to map the surface of Mars from a combination of rover and satellite imagery.
Reconstructing terrain millions of kilometres away, from imperfect and irregularly spaced data, demanded a level of geometric rigor that most commercial photogrammetry tools never have to meet. That same engine, adapted for handheld 360 cameras and smartphones, now produces engineering-grade 3D models of construction sites, geotechnical slopes, and infrastructure assets.
Where accuracy comes from
For casual or visual use, raw photogrammetry output is often good enough as-is. For engineering and construction work, accuracy has to be measurable and repeatable. Two things drive that.
Scale and control
Physical targets give the pipeline a known reference for real-world scale, cutting drift error significantly. Surveyed control points bring absolute position error down to survey-grade tolerances.
Surface type
High-texture materials like concrete or timber give the matching algorithm plenty to work with. Reflective surfaces and flat painted walls give it very little.
Where this fits in an AEC or geotech workflow
For site documentation, inspection, and progress capture, this kind of camera-based capture covers far more ground per hour than static laser scanning, without needing a specialist operator or dedicated hardware. Static total stations and terrestrial laser scanners remain the reference standard for the highest-accuracy survey work, but for the day-to-day job of recording a site as it actually is, video-based capture from a 360 camera or smartphone is increasingly the faster, cheaper option, especially when the output needs to plug directly into tools like QGIS, ArcGIS, AutoCAD, or Revit.
That's the process, end to end: from a walkthrough video, to a point cloud, to a measurable, shareable 3D model, with no scanning specialist or laser hardware required.
Let's talk about your next project
Get in touch to talk through a specific site or project. We'll walk your team through the capture, the accuracy, and the handover into your existing workflow.
Phone
+64 21 611 906
Based in
Wellington, New Zealand