Matrice 4E - 13.2 ha Survey Mapping | Nadir vs Smart Oblique 25° | GSD mismatch & GCP layout advice needed

[Drone BPL]

Hello everyone, how are you guys doing?

First time posting here as I’m preparing my first serious survey-grade mapping project with the Matrice 4E and would really greatly appreciate feedback from those with more field experience. I tried to sum up all of the details as concisely as possible, so please bear with me! My goal is to do this "by the books" and skip the guessing work. Thank you so much )

Project Overview​

• Waste water treatment facility (planned)
• Area: 13.2 ha (encompassing a bit more than the cadastral map just for the sake leeway/accuracy in post)
• Mixed terrain: mostly flat ground along with some hilly terrain, a few vertical objects, trees, and a small body of water in the center. Photo attached.
• Goal: high-quality orthomosaic + DSM + dense point cloud (maybe even 3D export)
• Deliverable standard: near survey-grade or maximum possible (cm-level accuracy)

---

Flight Parameters (Current Plan)​

Altitude:​

• 57.3 m AGL

Overlap:​

• 85% front
• 80% side
• Estimated flight time: ~50 min (single grid)

GSD:​

• Nadir: 1.54 cm/px
• Smart Oblique (25° gimbal): 1.70 cm/px

RTK:​

• Enabled and Fixed.
• Logging RINEX

---

Main Questions​

Smart Oblique vs Two Separate Flights​

Would you:

A) Run Smart Oblique at 25° in one combined mission
OR
B) Fly two separate missions:

• One pure nadir (-90°)
• One oblique grid (-65° to -70°)

Is there a geometric advantage in separating them for block strength, or is Smart Oblique sufficient for a 13.2 ha site? Or would you do it in a different way, according to your experience? As to you, does smart oblique keep the data distortion free, or is there an issue while processing, in terms of scale or anything else? NADIR + oblique.

I use Agisoft Metashape Pro.

---

GSD Difference (1.54 vs 1.70 cm)​

The ~10% GSD difference between nadir and oblique -
could that introduce any scale inconsistency or vertical bias?

My understanding is that bundle adjustment normalizes scale during processing, so this should not cause measurable distortion — but I’d appreciate confirmation from those who’ve tested it.

---

.
Optimal Oblique Angle​

Currently considering:

• 25° oblique (≈ -65° gimbal pitch)

Would you recommend:

• 20° for terrain stability?
• 25° for balanced geometry?
• 30° for stronger intersection angles?

Site has:

• Some vertical edges
• Curbs
• Tree boundaries
• Small water body

Goal is improved Z stability and edge reconstruction, not full 3D facade modeling.

---

GCP Plan​

Planned:

• 5 GCPs (4 perimeter + 1 center)
• 1–2 checkpoints

Target size:

• 30 cm x 30 cm
• Black/white high contrast
• Expected pixel size:
  
  • ~19 px in nadir
  • ~17 px in oblique

Is 30 cm sufficient at 1.5 cm GSD, or would you go 40 cm for safer marking? I'll use these checkpoints for comparison with my colleagues terrestrial measurements.

---

Checkpoints on Curbs​

Some existing survey marks are located on curbs.

Concern:

• With pure nadir, curb edge might not reconstruct cleanly.
• Could introduce marking bias if model smooths the edge.

Would oblique imagery significantly improve curb definition for checkpoint reliability?

Or would you recommend placing checkpoints on fully flat pavement instead?

---

Trees & Water​

There is a small body of water in the middle and tree clusters

So I expect:

• Weak reconstruction over water (acceptable)
• Possible canopy gaps

Would Smart Oblique at 25° meaningfully improve canopy density, or is it marginal for this scale?
Did you ever use a ND filter, did it help? Or did it make the shutter speed alot worse?

---

Processing Constraint​

Currently limited to 8GB VRAM, which may struggle with dense cloud processing at this overlap and image count.


I know it's a stretch and taking it for granted but I have to ask:

Wondered if anyone would be willing to help me process the dataset. BUT, it would be 4600 images...

I would be EXTREMELY grateful. Planning to upgrade workstation soon if all goes to plan!

---

My Objective​

This is my first major professional project and I want to avoid rookie mistakes or guesswork as much as possible. And deliver accurate results that I can stand behind.

Any critique, parameter adjustments, or workflow advice is highly appreciated.


Thank you everyone, in advance. Very grateful to discuss with all of you!
Wish everyone a productive week!

Mladen

 

[jaja6009]

You can run this with Smart Oblique, nadir with altitude optimization (1 flight line of oblique from outside to center) or even with double grid oblique.
If you run double grid oblique you do not need the overlap listed since oblique images contain so much more of the scene in each image. I run double grid oblique 30 degrees with good results with 70/70 overlap. This will help if you need a 3D model and will now have less images to process.

GCP's will help to tighten up your data, so this is a good idea. Make sure they are not too close to the edges.
I would add more check points. I would do a minimum of 5 and possibly 10. Include check points at high and low areas and like your GCPs spread them out evenly.
I use 2ft x 2ft control points and have zero issue seeing and marking them. On projects demanding proven accuracy I do not use items in the area, only dedicated control point targets to ensure accurate marking.
Check points are how you prove your absolute accuracy, so do not have too few of them.

The water will most likely reconstruct poorly and you will most likely have to clean those points up manually in the dense cloud.

Do NOT use an ND filter. I have actually had very good sharp, focussed images with the current Matrice 4 E mapping settings.
Best results will come on a cloudy day with no shadows, but you would still want to map midday. The clouds will diffuse the lighting and remove shadows. I use Aperture Priority 4. The camera then adjusts the ISO and Shutter Speed. You'll notice that when the mission starts the drone will fly to the start point, position the camera angle autofocus once and then lock in that focus for the rest of the mission.

How much RAM does your computer have? This is where if you do not have enough your processing will fail or run forever.

I would also take a look at this article. It helps cleaning up the initial sparse cloud.

USGS Publications Warehouse

Structure from motion (SFM) has become an integral technique in coastal change assessment; the U.S. Geological Survey (USGS) used Agisoft Metashape Professional Edition photogrammetry software to develop a workflow that processes coastline aerial imagery collected in response to storms since...

pubs.usgs.gov

Make sure you know what coordinate system datums:

• Your drone collects images in
• Your GNSS Rover Collects in
• Your RTK Base Sends in
• Your client wants

Agisoft is excellent in regards to flexibility with this. Make sure you have the correct geoid undulation tiff if you need orthometric elevations and it is defined correctly in Agisoft.
I have good results in collecting control points in the same geographic system that the drone receives corrections in using ellipsoid elevations, and then allow Agisoft to apply the geoid undulation.

If you need a DTM, Agisoft does an okay job classifying points in my experience. Cloud Compare can do a better job if you know how to use it properly.

If you create contours though, I would not use Agisoft. I would thin out the dense cloud or even better make a grid of points with a point per 5 meters and then create contours from this. Contours look terrible from ultra dense point clouds with noisy ground.


Good luck on this project!