Site Reconstruction by Camera Matching
Course Description
Site Reconstruction teaches you to rebuild a physical location in 3D from satellite imagery and OSINT imagery — photos and video. You start by anchoring a georeferenced world in Blender from satellite imagery, then match your camera to the source photos using vanishing-point perspective solving (the fSpy process), and finally model the site from the matched camera and verify the result by photomatching against the source frames.
The course concludes with a final project: reconstruct a house from satellite imagery and multiple photos. A deliberately generic scenario, built so the same pipeline translates to any case you work on.
The content is developed from a live 8-week curriculum (Oxford + Dutch Police pilot) and composes existing KB reference pages — camera matching, measurements, geometric reconstruction — into a single teaching progression.
Prerequisites
- Foundation: Blender for Visual Investigation (course 00) — completed or equivalent
- Comfort navigating object mode, adding/moving objects, basic modifier stacks
- No prior investigation experience required
- A working install of Blender (4.x) with the fSpy importer and BlenderGIS add-ons
Modules
| Module | Title | Est. Time | Type |
|---|---|---|---|
| 01 | What is Site Reconstruction? | 30 min | Foundations |
| 02 | Establishing the Satellite Baseline | 60 min | Pipeline |
| — | Measurements, Scale & Dimension (shared reference module) | — | Reference |
| 03 | Camera Matching: Perspective Solving | 60 min | Pipeline |
| — | Camera Matching (shared reference module) | — | Reference |
| 04 | Modeling from Matched Photos | 60 min | Pipeline |
| — | Object Reconstruction (shared reference page) | — | Reference |
| 05 | Tools & Addon Ecosystem | 30 min | Tools |
| 06 | Investigation Context: Scale, Distortion, Documentation | 30 min | Methodology |
| 07 | Final Project: Build a House from Photos | 120 min | Apply |
Total estimated time: ~7 hours
Learning Objectives
By the end of this course you will be able to:
- Explain the two anchors of site reconstruction — a georeferenced world and a matched camera — and when the technique is (and isn't) appropriate
- Establish a 1:1 georeferenced baseline from satellite imagery in Blender
- Perform camera matching from a single photo using vanishing points / perspective solving (the fSpy process)
- Model a site from the matched camera and verify the model by photomatching source frames
- Choose the right satellite/GIS tool and matching tool for a given source set
- Document a reconstruction (scale hierarchy, distortion, methodology) so it can be scrutinized
- Complete the final project: reconstruct a house from satellite imagery and multiple photos
How to Use This Course
- Work through the numbered modules in order.
- Modules 02–04 each embed the corresponding video walkthrough from the BVI series; watch the video, then read the module notes.
- Items without a module number (Measurements, Camera Matching, Object Reconstruction) are shared KB reference modules woven into the sidebar flow — read them at the point they are introduced and keep them open for reference.
- The Final Project (Module 07) is done in Blender with a prepackaged asset set; no case-specific knowledge is required.
KB Reference Pages Used
| KB page | Used in |
|---|---|
| Camera Matching | 01, 03 |
| Measurements, Scale & Dimension | 02 |
| Camera and Perspective | 01, 03 |
| Analyzing Footage | 03 |
| Scale Estimation | 04, 06 |
| Geometric Reconstruction | 01, 04 (extension) |
| Scene Management | 04 |
| fSpy add-on / Perspective Plotter | 03, 05 |
| BlenderGIS / Blosm / QGIS / Google Earth Pro | 02, 05 |
Out of Scope
- Object/evidence reconstruction (e.g. a specific D-30SN) — separate course track
- Camera tracking (reconstructing moving cameras from video) — separate course
- Full urban block-out — see the extension pointer in Module 04 (environmental reconstruction)