Case study · Blair Street 2
One planned flight over a single Summerland property. Six thousand-odd images, each one locked to a corrected position. We processed the whole job in-house, and that one capture became a 3D house mesh, a classified point cloud, a zoomable orthomosaic, and a rendered flythrough. Here's how it actually went, stage by stage.
The capture, stage by stage
No re-flights, nothing stitched together from separate jobs. Everything below follows a single capture from the first phone call through to the walkthrough. Every figure on this page is measured, not estimated.
A 5.3-acre Summerland property: house, driveway, grounds, and the vineyard hillside rising behind it. We wanted to put the full pipeline through its paces on one site. Capture it once, then see how many separate deliverables a single good flight can support. One property, every output, no shortcuts.
The site: 5.3 acres · house, grounds & vineyard slope · Summerland BC
We planned the mission in NAV Drone against current Canadian aviation rules before anything left the ground. Overlap, altitude, and flight lines were set to carry both the wide property context and tile-level roof detail in one pass. Airspace got checked at planning, not on the day.
Planned in NAV Drone · flight lines set for full coverage and close detail
One planned flight produced 6,052 images, each frame georeferenced as it was taken on an RTK positioning workflow. The result on the day was 100% RTK FIX. Not float, not a single dropped frame. That's the difference between a pretty picture and data you can actually measure from later.
6,052 RTK-fixed images · 100% RTK FIX · one flight
The whole set ran through DJI Terra and the QGIS and Blender pipeline on our own machines. That's roughly 25 hours across 46 processing blocks, handled by the same pilot who flew the site. Nothing waited in a third-party queue, and outputs came back georeferenced in NAD83(CSRS).
≈25 h · 46 blocks · DJI Terra → QGIS → Blender · NAD83(CSRS)
The finished model lands within 3.6 cm RMSE of where the ground actually is. That's a verified figure from the processing report, not a number we hoped for. That accuracy is what makes the same capture trustworthy as a measurement source, not just something nice to look at.
3.6 cm georeferencing RMSE · verified, not estimated
The outputs landed in a shared folder with interactive 3D you can open in a browser, plus a walkthrough of what the data shows. Everything below this point is a real deliverable from this one capture. Load each one and look around.
Shared client folder · interactive 3D · walkthrough of the data
What came out of it
These are the real outputs from the flight above. It's the interactive 3D a client gets, loaded straight from our own Blair Street capture. All of it came from that one flight.
A photogrammetry mesh of the house and grounds, built from the same 6,052 images. Drag to orbit, scroll to zoom. This came from that one capture.
An eight-million-point web preview of the LAS deliverable. Switch between true colour, ground / above-ground classification, and elevation. This came from that one capture.
The full 5.3-acre property as a top-down map you can zoom into 0.3 cm/px delivered pixel detail. That's close enough to read roof tiles. This came from that one capture.
A 57-second camera path rendered straight through the photogrammetry twin. It's marketing footage you can't get from a flat photo. This came from that one capture.
How every job runs
Blair Street wasn't a one-off setup. It's the standard pipeline, run end to end. Here are the three stages that make a single capture worth this much.
RTK positioning on every flight, planned in NAV Drone, with GCP targets placed where the job calls for independent ground reference. That's what lets the data hold up as a measurement source, not just a photo.
What you walk away with: data captured to a documented standard
DJI Terra through the QGIS and Blender pipeline on our own machines. Nothing waits in someone else's queue, and georeferenced outputs come back in NAD83(CSRS).
What you walk away with: georeferenced outputs in NAD83(CSRS)
A shared client folder, interactive 3D where the job calls for it, and a walkthrough of what the data actually shows. It works exactly like the deliverables on this page.
What you walk away with: deliverables you can open, share, and act on
Questions & answers
No. Blair Street 2 is our own demonstration capture, shown exactly as a client would receive it. Everything published on this site is our own demonstration capture. Client work appears publicly only with written permission.
Yes. The house mesh, the classified point cloud, the zoomable orthomosaic, and the flythrough all came from the same 6,052 images captured in a single planned flight. No re-flights, no separate jobs stitched together.
It's the georeferencing root-mean-square error from the processing report. In plain terms, it's how closely the finished model lines up with where the ground actually is. It's a measured figure from this capture, not an estimate, and it's what makes the orthomosaic trustworthy as a measurement source.
Request a quote
Tell us about your property and what you need to see. Every project starts with a short scoping conversation about what you need, what we'd fly, and exactly what you'll receive.