Guide

Cowboy guide to drone mapping: a pragmatic field guide

A field-tested guide to drone mapping. The 5-photo rule, 2D, 3D Area and 3D Object modes, circle flights around buildings, GSD, overlap, gimbal angle, GCPs, RTK, doming, dataset hygiene and how to avoid expensive reflights.

By Lars LindvedUpdated 10 October 2026

The short version

Choose the mode by how you flew: a grid gives 2D or 3D Area, close circles around a building give 3D Object.
The 5-photo rule: every spot you want in the result must be seen in at least 5 photos.
Perfect missions are nice, not required. Circles can be flown by hand.
Do not mix -90° photos and angled photos in a 2D job.
A double grid is cheap insurance against data gaps.
Bring in a certified surveyor if the result needs legal validity.

Disclaimer: mapping is not certified surveying

Let us start with the important part: I am not a licensed surveyor, and this is not a university textbook. This guide is built on my own field experience, opinions, mistakes and successes over the years.

In this guide, I use mapping and certified surveying as a practical distinction. The words mapping and surveying are not used in exactly the same way in every country, profession or company. It is not a universal legal definition, and there are many shades of grey between the two. We will not cover all of them here.

When I say mapping, I mean creating a useful visual and geographic model of a roof, construction site, stockpile, field or landscape. Mapping is the eye in the sky. You want to see what is there, track progress, find damage or plan the next move. The standard is whether the model answers the practical question, not whether it is a certified survey.

Certified surveying is a different discipline. If the result must establish boundaries, support construction set-out, prove compliance or settle a dispute, it is surveyor territory. RTK, GCPs and an expensive drone are tools. They are not a magic certificate.

Take a roof inspection. If the question is: Is there a hole, and where should someone look? A visual model may be exactly the right solution. If the question is: Can we prove the building is higher than allowed? That is a different job, and a certified surveyor should be involved.

It is like choosing from a garage with both a supercar and a mini-van. A supercar is not better than a mini-van for buying groceries. Once the groceries are in the kitchen, nobody can tell which car delivered them. Both cars can drive around a racetrack, but if the goal is a certified lap time, the supercar is the right tool.

The right drone is not the most expensive one. It is the one that matches the question.

We also do not dive into aviation law, A1/A3 certificates, geo zones or local permits here. Rules vary depending on where you fly, and it is always your responsibility as pilot to stay on the right side of the law.

The purpose of this guide is simple: To give you a pragmatic field guide to drone mapping. How do you get fantastic, usable orthophotos and data out of your drone without wasting time or wrecking expensive gear?

Concepts you need to know without the theory shock

Drone mapping or photogrammetry is simply taking overlapping photos from the air and letting a computer stitch them into 2D or 3D models. Here are the key parameters:

GSD (Ground Sampling Distance): A resolution measure. It tells you how much ground one pixel covers, for example 1.5 cm/pixel. Lower is sharper, but you must fly lower and take many more photos.
Front and side overlap: How much photos overlap in flight direction and sideways, typically 70 to 80 percent.
Gimbal angle: The angle your camera points, DJI style: -90° looks straight down (nadir), perfect for flat 2D maps. Angled photos (oblique, for example -45° to -80°) look forward and down, essential for capturing sides of a chimney or building for 3D.
Single grid vs. cross grid: A single grid is the classic lawn mower back and forth. A cross grid (double grid) adds a second pass at 90° to the first. It takes longer but closes gaps in your data.
Orthophoto (2D): A flat, measurable aerial photo corrected for lens and perspective distortion.
Elevation model (DSM / DTM): A digital map. DSM (surface) includes trees and buildings, while DTM (terrain) shows the bare earth.
Point cloud and 3D mesh: A 3D model built from millions of coordinated points.

You can compute GSD, overlap and flight altitude in our free flight planner.

Photogrammetry is not magic: the 5-photo rule

Many people think drone mapping is about perfect missions, apps and expensive gear. After years in the field I have learned it comes down to one rule: every spot you want in the result must be seen in at least 5 photos, taken from slightly different positions. That is what 80 percent overlap means in practice: as you move, each spot stays in frame for at least 5 shots.

A mission app is a convenient way to keep that rule over a large area. A calm pilot flying by hand can keep it too, and so can a person walking with the drone in their hands. What the camera never saw, the computer cannot invent: under the eaves, behind a chimney, inside a balcony. If it is missing in the model, it was missing in the photos.

Three flights, three modes

When you upload, you choose 2D, 3D Area or 3D Object. Choose by how you flew, not the other way round.

	2D	3D Area	3D Object
For	Maps, roofs from above, sites	Forest, construction site, neighbourhood	One building or structure
Flight	Single grid	Cross grid: a second pass at 90° to the first (the Cross grid option in our planner)	Close roof grid + circles + folded-arms photos
Gimbal	-90°	-80°	Roof grid -65° to -80° (-75° in our planner)
Distance	Height for about 1 cm/px: about 28 m on Lito X1, about 55 m on Mini 5 Pro	Whatever the job needs	No more than about 10 m from the object. Planner altitude = roof height + distance
Expect	A sharp map	A fair map and a fair model, neither perfect. It is a compromise	The best model, but no map

2D

For: Maps, roofs from above, sites
Flight: Single grid
Gimbal: -90°
Distance: Height for about 1 cm/px: about 28 m on Lito X1, about 55 m on Mini 5 Pro
Expect: A sharp map

3D Area

For: Forest, construction site, neighbourhood
Flight: Cross grid: a second pass at 90° to the first (the Cross grid option in our planner)
Gimbal: -80°
Distance: Whatever the job needs
Expect: A fair map and a fair model, neither perfect. It is a compromise

3D Object

For: One building or structure
Flight: Close roof grid + circles + folded-arms photos
Gimbal: Roof grid -65° to -80° (-75° in our planner)
Distance: No more than about 10 m from the object. Planner altitude = roof height + distance
Expect: The best model, but no map

Need both a 3D model and a map of a building? Fly the map as a separate flight at -90° and upload it as its own 2D job. Do not mix the two photo sets, and you never pay twice for the same photo.

Circle flights around buildings

By hand is fine, and often better. Low circles near trees, wires, lamp posts and scaffolding are where automated orbits get risky. Obstacle sensing does not reliably see a thin branch or a wire, but you do.

Fly slowly in Cine mode.
Interval shooting every 2 s.
Keep the middle of the facade in the middle of the frame.
At least 36 photos per circle, one every 10°. Tight on the building (closer than about 6 m to the facade): 48 to 72.
Close-ups need a bridge: flying close to details? Fly a circle at medium distance first, so the computer can connect the close-ups to the rest. Without the bridge, close-ups end up as loose pieces or get left out entirely.

Pass	Height	Gimbal
High circle	Roof + 5 to 8 m	-45° (roof edge down to the base)
Low circle	Eaves or mid-facade	-10° to -20°
Folded-arms walk	Ground	Tilted up as far as it goes (+35° on Lito X1)
Roof grid	Roof + up to 10 m	-65° to -80°

The folded-arms walk

Fold the arms, motors off, interval shooting on, gimbal tilted up. Stand 3 to 5 m from the wall and take one big step (about 1 m) per photo. That keeps every spot in at least 5 photos. Upload these into the same 3D Object job as your circles.

Want automation?

Mini 5 Pro: a POI orbit in DJI Fly with timed shot running.
Lito X1: check whether yours can do the same.
Waypoint orbit: plan it in a third-party planner and import the KMZ into DJI Fly.
Enterprise drones: DJI Smart 3D Capture (DJI Pilot 2) flies a rough model first, then hugs the building layer by layer. It is the professional version of the same idea, and it still lives by the 5-photo rule.

The myth of GCPs, RTK and the bowl effect

Every time you mention drone mapping, someone shouts: You need GCPs and RTK or your map is useless. Let us debunk that myth.

Relative vs. absolute accuracy

Absolute accuracy: Is my map millimetre perfect on the global cadastre? Spoiler: This is the unicorn of photogrammetry. Even Google Earth satellite images are rarely millimetre precise.
Relative accuracy: Is a 10 metre wall on my map also 10 metres in reality?

What is RTK and do I need it?

RTK (Real-Time Kinematic) is GPS on steroids. While normal drone GPS can be a few metres off, an RTK drone constantly talks to a base station or network to know its exact position in the air down to centimetres.

What does it do better? It gives each image an absolute world reference for camera position and can reduce the need for physical GCPs. That is not the same as guaranteeing absolute accuracy in the finished model.
Why is it often overkill? It costs a fortune, both a pricier drone and often an expensive mobile network subscription. If you just need to measure a roof, standard GPS and relative accuracy is often enough for a rough estimate. For measurements used in billing or contracts, check them with your own control measurements.
When can it be needed? If your map must align with existing survey data, municipal utility maps, or if you fly the same site every week and want maps from week 1, 2 and 3 to stack precisely automatically.

Why use GCPs? (Ground Control Points)

GCPs are surveyed markers on the ground that anchor the model to known coordinates. They can be especially important when absolute height matters in an elevation model. We almost never use them ourselves, because it is often total overkill and wastes 45 minutes in the field.

But height is the annoying part. Even expensive RTK drones can still leave absolute height errors in the finished model. If you fly over completely flat terrain with low overlap and the camera straight down, you can get the bowl effect (doming). The computer miscalculates and thinks the ground curves like a soup bowl.

Dataset hygiene: do not mix pears and bananas

A classic mistake is to fly a building thin with both straight-down (-90°) and angled photos, dump all 500 images in one folder and expect the computer to magically spit out both a perfect 2D map and a 3D model. It does not work. The rule depends on the mode:

Dataset A (2D and 3D Area): Keep -90° photos and angled photos apart. For a flat map the camera must look straight down, so a 2D job takes only -90° photos. If you feed the engine photos looking into a wall or toward the horizon, it chokes trying to force them into 2D. For 3D Area the whole cross grid at -80° is one set: fly all of it at the same angle and upload it together.
Dataset B (3D Object): Mixing angles is the whole point. The close roof grid, the circles and the folded-arms photos all go into the same 3D Object job, because the computer needs to see every spot from many directions to build facades and eaves. What you keep out is the -90° map flight: that is its own 2D job.

Risk, gear choice and the expensive refly

In drone business the math for all mapping is: Flight time plus number of images equals cost. Batteries, time on site and heavy server processing cost money. But before you cut everything to the bone to save money, know the real pitfalls:

1. Risk factor (why use a cannon for sparrows?)

Every flight is a risk. Birds, wind gusts and invisible cables exist. Why risk smashing an enterprise drone with a 150,000 kr. camera over a construction site if the job is just a 2D orthophoto at 2 cm/pixel? A small DJI Mini 5 Pro can deliver a result few people, and even fewer clients, can tell apart. Choose gear by risk.

2. The refly penalty

Mathematically double flight time and double images sounds like a budget killer. But the most expensive thing is not taking 50 photos too many. The single most expensive thing is sitting at the office, discovering overlap was too low, and being forced to pack the car and drive out to refly the job. That is why we often fly a double grid even though we think a single grid is enough. It is cheap insurance.

3. The scale paradox and geography

What is smart depends entirely on job size and location:

The small roof (200 m2): Here it takes 2 minutes extra to take twice as many photos. The drone still runs on the same battery. Overkill costs nothing but secures the data.
The big field (80 ha): Here double flight time is a disaster. You go from 1 battery to suddenly needing 4 batteries and field charging. Here you cut to the bone and fly minimal overlap.
Geographic distance: Is the job 2 hours away by car? Then always fly with a massive safety margin of extra photos. You do not want to drive 2 hours back tomorrow.

Real world examples

Here are 3 everyday scenarios:

1. Roof mapping (both 2D and 3D)

Challenge: Sharp angles and shiny roof surfaces. Area is small.
Cowboy advice: Roof is small so battery is no issue. Two flights, two jobs. A -90° grid at about 1 cm/px goes in as 2D. The close roof grid, the circles and the folded-arms photos go in as 3D Object.

2. Construction site (residential construction)

Challenge: Weekly status on earth piles. Must be fast.
Cowboy advice: Standard cross grid (double grid) at 40 to 60 metres with a reliable drone. Want 3D? Fly the cross grid at -80° and upload it as 3D Area. If a map is enough, fly it at -90° and upload it as 2D. You do not need millimetre precision to see if earth has moved.

3. Agriculture (RGB mapping)

Challenge: Large areas where you want to see variation or document progress with ordinary RGB images.
Cowboy advice: Use a coherent RGB dataset with the resolution and overlap the job requires. Plan for batteries, terrain and local rules. You need variation in the field, not to count ladybugs.

Summary

Drone mapping starts with a reliable drone, a sensible safety margin and clean datasets. Adjust your margin by whether you are 10 minutes or 2 hours from home, and keep -90° photos and angled photos apart. That gives you a better chance of getting the data right in the field.

Frequently asked questions

How many photos should I plan for?

The count is driven by GSD and area, not a fixed number. At 70 to 80 percent overlap an ordinary roof job needs roughly 150 to 300 images, while a medium construction site easily passes 500 to 1,000. Rule of thumb: halve the GSD and the image count roughly quadruples, because each image then covers a quarter of the ground.

What accuracy can I expect without GCPs?

Measurements inside the model are useful for most jobs. If you need to know how close they are, measure a few control points in the field and compare. The absolute position on the world map can be off by several metres. If you need absolute accuracy at centimetre level, GCPs and a surveyor are the right route, not more photos.

How many GCPs do I need if I use them?

Plan for at least 5 well-spread points, ideally one near the centre for vertical control, and more as the area grows. Every point must be surveyed to centimetres. If they are clustered in one corner they will not anchor the model, and you have wasted the trip.

Do I need a mission app for 3D?

No. The 5-photo rule matters, not the app: every spot in at least 5 photos. Circles can be flown by hand, and near obstacles that is often the safer choice.

How many photos per circle?

At least 36 (one every 10°). Closer than about 6 m to the facade: 48 to 72.

Can I use the same photos for 2D and 3D Object?

No. Fly the -90° map as its own flight and upload it as a 2D job, and the close grid, circles and folded-arms photos as a 3D Object job. Separate sets, separate jobs, and no photo is paid for twice.

When is a double grid cheaper than a refly?

Rule of thumb: if the drive to the site is longer than the flight itself, fly the double grid. On a small roof the extra pass costs almost no battery, while an 80 ha field can mean several batteries and field charging. On large fields with a tight battery budget, a single grid at 75/70 overlap is the pragmatic choice.

Ready to test your images?

Upload your images and get a GIS-ready orthophoto without a subscription, or contact us if the job is large or needs a specific deliverable.

Create account and upload Contact us