Case Study #1 · Raw-photo reconstruction

547 raw drone photos in.
A full map set out, in 44.6 minutes.

No desktop photogrammetry software. The raw geotagged frames went up straight off a DJI RTK aircraft, and DroneFlow.Ai reconstructed the orthomosaic, elevation models, contours, point cloud and 3D model. Below: the output, the flight it came from, and what a flight like this can and cannot establish.

This is the same ground as Case Study #1, which starts from a processed DJI Terra export instead. Same flight, two pipelines.

547
RAW PHOTOS UPLOADED

200 nadir + 347 oblique, geotagged in-EXIF. Nothing else was supplied.

44.6 min
RECONSTRUCTION TIME

Upload to finished deliverable set, at the fast preset.

8.13M
POINT-CLOUD POINTS BUILT

8,132,721 points, LAS 1.4 — read from the file header.

6
DELIVERABLE TYPES

Each file was opened and parsed before this page was written.

The orthomosaic DroneFlow built from the raw frames

Orthomosaic of the Case Study #1 materials yard reconstructed by DroneFlow.Ai from 547 raw DJI photos: a highway junction, two rail lines, stacked culvert stock, aggregate piles, a concrete pad, active equipment and a retention pond
A materials-handling site: highway frontage, twin rail lines, aggregate stockpiles, stacked culvert stock, active equipment and a retention pond — stitched from the uploaded photos at 5 cm/pixel. The actual output, downsampled only for the web. It covers about 50 acres, against the 32.6 acres the delivered Terra product covers: the automatic boundary took in the junction, the second rail line and the woods south of the tracks. Same flight, wider footprint — the two maps are not pixel-for-pixel comparable.

What was flown

Capture parameters for the flight
Aircraft / cameraDJI M4E · 20.9 MP (5280×3956)
Frames547 — 200 nadir + 347 oblique
Flight altitude~51 m AGL
Native GSD~1.3 cm/px
PositioningDJI RTK, geotagged in-EXIF
Ground control / checkpointsNone on this flight

What DroneFlow produced

Deliverables reconstructed from the raw frames
Orthomosaic11,109×7,396 px · 5 cm/px
DSM & DTM10 cm/px · 44.5 m relief
Contours7,615 lines · 1 ft interval
Point cloud8,132,721 pts · LAS 1.4
3D modelTextured mesh + 10,134 web tiles
Coordinate systemWGS84 / UTM 17N (EPSG:32617)

What we measured — and what we didn't

Every deliverable above was opened and validated by the automated pipeline self-checks before this page was written. The two notes below are not caveats added afterwards: they are the boundary of what a single RTK flight with no surveyed control can establish at all.

  • PASSAll 6 deliverable types produced from raw photos — orthomosaic, DSM, DEM, contours, point cloud and textured 3D model — and each file opened and parsed cleanly.
  • PASSAll 547 frames georeferenced; the browser 3D tileset built 10,134 payload-sized tiles (median 14 KB), the size range that streams without stalling.
  • NOTEAbsolute accuracy is anchored by the aircraft's onboard RTK positions (per-photo σ ~8 cm horizontal / ~15 cm vertical), not independently verified — this flight carried no surveyed ground control or checkpoints, so no residuals can be reported. For design or record work, add checkpoints and a licensed review.
  • NOTEThis map was built at the fast processing preset. The point cloud (8.1 M points, 5 cm ortho) is intentionally lighter than a premium desktop engine would produce from the same frames — a deliberate speed/detail trade, tunable per project.

Accuracy depends on the aircraft, camera, altitude, flight geometry, image overlap and quality, RTK/PPK correction, ground control, checkpoints, terrain, vegetation, coordinate system, and processing configuration. Centimeter-level results may be achievable with appropriate RTK, PPK, surveyed ground control, checkpoints, and qualified validation. DroneFlow.Ai outputs do not constitute a licensed land survey unless reviewed and certified by an appropriately licensed professional.

Have raw drone photos? Skip the desktop software.
Upload the frames and DroneFlow builds the map. 7-day free trial — no credit card required.
See pricing