Making Great Maps:

The Complete Guide to Professional Mapping

with DroneDeploy
By Adam Carp
Forward
Welcome to ‘Making Great Maps: The Complete
Guide to Professional Mapping with DroneDeploy’.
This guide is designed to be the ‘one place’ where you can find the
majority of your questions about mapping with DroneDeploy answered.
It can either be read from start to finish, giving you a wide baseline
knowledge of connected topics or, you can pick out individual sections
to read case-by-case as you need them. You can skip a lot of real-life trial
and error by reading this document in detail.

Here is a summary of the journey we’ll take you on:

We start this e-book by describing the basic processes behind Map Engine.
Understanding the basics behind structure-from-motion photogrammetry
processing allows you to create datasets that are tailored towards success.
Next, we’ll walk you through each of the most important flight modes and
describe their use cases and the best techniques to keep in mind when using
them. Once you’ve chosen your flight mode, refining those settings using the
‘Advanced’ panel can be important to assure a quality result, so we’ll describe
all of the parameters available in that section as well.

The physical aspects of the subject, such as elevation change, homogeneity,

and reflectivity, will often require further adjustments to your flight plan.
We’ll describe those common complications and how you can further refine
your plan to best avoid a negative impact from these complications. The final
pieces to factor in when designing a flight plan are your resolution and
accuracy goals. This includes all of the considerations needed to successfully
map using Ground Control Points.

At that point, you’ll be set to complete the design of your flight plan. We’ll
then describe suggested workflows and checklists for pre-flight preparation
and on-site flying. This ensures that you can avoid hiccups in the field that
might impact the efficient use of your time.

To ensure that your hard work pays off, we’ll explain how to correctly
process your data in the right modes. Finally, we’ll then give you the tools
and knowledge you’ll need in order to quality check your completed map
and to continue building on your photogrammetry and mapping knowledge
for the future.

As always - Happy Mapping!

 05 Tie In Your Accuracy And
Resolution Goals

Absolute vs Relative Accuracy . . . . . . . . . . . . . . . . . . . . . . . . 26

Table of
Increasing Accuracy Without Adding GCPs . . . . . . . . . . . 27
High Precision Mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
GCPs Or No GCPs? . . . . . . . . . . . . . . . . . . . . . . . . . 28

Contents
Checkpoints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
GCP Placement Rules . . . . . . . . . . . . . . . . . . . . . . 29
Tagging Your GCPs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
Interpret Your Accuracy Report Data Correctly . . . . . . 31
Resolution And Altitude. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

01 Understand How Structure From

Motion Photogrammetry Works 06 Create A Flight Plan

Balance and Experimentation . . . . . . . . . . . . . . . . . . . . . . . . . 36

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 05

02 Choose The Correct Flight

Planning Modes
07 Bring Your ‘A’ Game To The Field!

Pre Flight Checklist . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

Non-Structures Mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 06 Best Practices In The Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
Structures Mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 07
POI’s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 09

08
Crosshatch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Process Quality Data In The
House & Roof, Precision 3D Model . . . . . . . . . . . . . . . . . . . . 11
Correct Mode
Linear Flight Plan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Live Map HD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
What Does A Good Image Look Like? . . . . . . . . . . . . . . . . . . 39
Manual Flight . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
What Do Bad Images Look Like? . . . . . . . . . . . . . . . . . . . . . . . 40
Map Engine Processing Requirements . . . . . . . . . . . . . . . . 41

03
Terrain Mode Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
Know The ‘Advanced’ Panel
Structures Mode Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

Automatic Settings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 Experimenting With Processing Modes . . . . . . . . . . . . . . . 42

Advanced Settings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

09
Offline Mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
Quality Check Your Map

Sanity Check . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

04 Consider How Your Subject Can

Complicate Things And How You
Alignment Check . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
Image Projections Check . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
Can Respond Image Density Check . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
Location Pin Image Check . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
Homogeny . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
Accuracy Check . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
Elevation Change . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

10
Intricate / Fine Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
Recipe Book for Common
POI’s And Difficult-To-See Angles . . . . . . . . . . . . . . . . . . .. . . 23
Project Goals
The Shape Of Your Subject Area . . . . . . . . . . . . . . . . . . . . . . . 24
2D Layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
Reflectivity And Lightning Conditions . . . . . . . . . . . . . . . . . 24
Plant Health . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
Moving Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3D Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
Elevation Layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
 Introduction
Capturing, processing, and analyzing accurate drone data can
be as easy as a few taps on your mobile device. Our mission is
to make mapping easy and accessible, even for projects with
a tough list of requirements and a difficult subject.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 4
 01
It is important to first understand the basic concept behind
drone photogrammetry. DroneDeploy’s Map Engine is our
in-house photogrammetry processing pipeline. It stitches your
image sets into complete maps and models using a process
called ‘structure-from-motion’. But how?
Understand
This process is similar to how we see with our eyes. Our two
How Structure- eyes give us two different perspectives of the same scene.
Over many years, our brain learns the distance between our
from-Motion eyes, and how that corresponds to the distance to an object is
based on seeing the same object with both eyes. Similarly, our
Photogrammetry photogrammetry pipeline works by matching visually distinctive
‘key-points’ that have been captured in multiple photos taken
Works from different locations. Each photo is like having another
‘eye-in-the-sky’ viewing the scene, and that allows us to create
a 3D reconstruction from photos.

Entire image sets are aligned into maps based upon similar
features found in multiple photos. The best maps have at least
8 images per single location.

Identifying and matching tiepoints across three photos

What Does Photogrammetry with DroneDeploy Truly

Come Down To?

The more matched, distinct objects across multiple photos we

can find, the more accurate the map becomes. These matches
we find are called ‘key-points’. In the example above, you can see
one key-points that have been matched between three photos.
Using DroneDeploy for flight ensures you take enough photos
to gather as many key-points needed to make a clear map.

Now that you know the basic concept behind the process
of photogrammetry, let’s learn about all of the tools at
your disposal.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 5
 02
Using the right flight planning modes is the first concrete step
you can take to put yourself on the path of making a successful
map. The different flight modes that you can use to capture and
process imagery in DroneDeploy each produce output that has
different strengths and weaknesses.
Choose the
Correct Flight Non-Structures Mode Flight

Planning Modes When you start off with a brand new flight template in
DroneDeploy, Structures mode will be off. When Structures
mode is off, your drone will capture a lawnmower pattern
(back and forth) capturing rows of nadir imagery ( images
taken from an overhead view with your drone’s camera
pointing straight down at the ground). This mode can
efficiently cover a large area.

Non-Structures mode flight pattern. The drone follows the green lines.

Non-Structures mode flights are best suited for big tracks of

land such as agricultural fields or site surveys. Non-Structures
mode datasets are well-optimized for homogeneous scenes
(where every image looks very similar to the next), especially
for agricultural fields.

While processing data captured in Non-Structures mode will

also produce 3D data products (point clouds and 3D models),
it’s best suited for accurate 2D layers and elevation profiles.
Non-Structures mode is best used for when you’d like to
efficiently cover a larger area without a relatively high image
count or without using a lot of batteries. Also, make sure
the scene doesn’t have any drastic elevation changes (cliffs/
buildings) or overhanging features (caves/arches).

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 6
 Within this section and later sections, we’ll go into more detail on how you can
properly address sites that require 3D models, for which a Non-Structures
mode flight will not suffice.

Structures Mode

Structures mode can be toggled on the left side of your flight planning
dashboard. It is optimized to reconstruct buildings, statues, objects, and
other objects in 3-dimensions. If you are looking for a compelling 3D model,
we would recommend starting with “Structures” mode.

In addition to the base set of nadir images (just like Non-Structures mode),
Structures mode also captures ‘oblique images’ (angled to capture the sides of
buildings or objects) from the perimeter of your mission plan, facing towards
the center of your subject. It does this while carefully not including the horizon
in the shots.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 7
 You can see the green waypoint lines around the perimeter in
the second photo on this page.

“The best
obliques
are taken
at a lower
altitute and
at a bit of a Notice the green waypoints around the perimeter of the flight. Structures mode
captures side angles of your subject, compared to Non-Structures mode, which is
just the lawnmower pattern.

horizontal The best obliques are usually taken at a lower altitude and at
a bit of a horizontal distance from your subject. Shooting too
distance” close to a building results in an extreme angle taken of the
vertical wall of that building.

As the photo angle becomes steeper and steeper, fewer and

fewer pixels are describing the same sized wall area, which
decreases the model’s quality and can even cause meltiness or
holes to form in the end-result.

Here’s an example of a shot that is too close to effectively help

with the vertical wall reconstruction:

When trying to reconstruct the close vertical wall of this brick building, this image is
taken too high up and too close.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 8
 If you shoot from farther away from the building and/or drop altitude, the side
of the building is seen more clearly. This results in more pixels describing the
same area now, which allows us to stitch the model with greater detail. Here’s
an example:

To ensure that all 360 degrees of your Structures images provide well-angled
coverage of vertical walls, it is important to properly frame your subject within
the orbit of your Structures flight.

POI - Point of Interest

An important concept regarding Structures mode is ‘Points of Interest’, or
‘POI’s’. POI’s are where the flight pattern and camera will need to focus on in
360 degrees to deliver a high-quality result. Almost any structure that you’d
like accurately modeled is a POI, for example, a large stockpile.

You might identify multiple points of interest in a single area, such as two
buildings on opposite sides of the site. For quality 3D results with multiple
POI’s within a single map, usually, combining multiple Structures mode flights
is required. More detail on that can be found below.

Why is it recommended to combine multiple POIs in a single map? This is

because Structures rotates 360 degrees around a singular POI.

In order to achieve the required 360 degrees of even coverage, with solid
vertical reconstruction, the POI needs to be centered within the oblique orbit
of photos, as seen below:

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 9
 If your primary structure (POI) doesn’t mostly fit into a circle of obliques, or
if you have multiple POI’s that don’t quite fit within a single orbit, it is likely
that the obliques will not capture adequate vertical data of those objects, and
they may not be reconstructed accurately. The 3D result will usually be better
than a Non-Structures mode flight would be, but it won’t produce the best
data possible.

So, what if you’re mapping an area with multiple points of interest, such as
a city block with a handful of high-rises? To remedy this, you can combine
multiple Structures mode flights to be processed into a single map by uploading
the images together. If possible, it is best to take off from the same location.
Minimizing your takeoff location and mapping altitude differences can help
keep the altitude data in your images consistent.

Set up each Structures flight to rotate around each important POI, and make
sure that each flight fully overlaps with the other flights. Here’s an example:

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 10
 Crosshatch Mode

Another fantastic option for a situation with multiple

POI’s is to use the crosshatch pattern from the Auto
Flights Modes app. Crosshatch doubles the number
of images that the lawnmower, Non Structures mode
style flights captures by intersecting each of those
rows with a perpendicular pass.

Because the image count is doubled, overlap is not

typically a large concern when using the Crosshatch
mode. Crosshatch mode only takes oblique, or
angled images. This increases the quality of the 3D
reconstruction of a large area with more than one
point of interest because this pattern does not center
around a particular point of interest but still does
A blanket of obliques provides a wide range of 3D coverage over provide a diversity of views over this large area.
a large area.

House & Roof, Precision

3D Model app
What if you want the ultimate single POI 3D
reconstruction that DroneDeploy can provide, or if
you need a detailed residential Roof Report? Look no
farther than the House & Roof, Precision 3D model
app, available on our App Market!

The top section of the House & Roof app is a

crosshatch pattern, exactly the same as described
in the previous section. This provides a dense and
wide baseline of views for the general area, which is
excellent for 3D modeling.

Not only does the Precision 3D app have the

crosshatch pattern, but it also orbits twice around
your subject, each at a different altitude. This is an
extra orbit compared to a Structures mode flight.

This makes the House & Roof app essentially a single

POI version of the Crosshatch mode or a higher-
detail version of the Structures mode.

It also has easy-to-use controls that automatically

set mapping altitude according to the altitude of
your point of interest and of obstacles in the area.
This helps to ensure that overlap is almost never a
concern with this flight mode.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 11
 What is the disadvantage of the Precision app?
Simply, time, SD card space, and batteries. It takes
a longer time in the air to cover the same area
compared to Structures mode flights. This effect
becomes increasingly prevalent as the mapped
area gets bigger, which is why the primary intention
of this app is to map a single structure and not
city blocks.

Linear Flight Plan

You may find yourself mapping a roadway, oil
pipeline, train tracks, or other linear infrastructure
that is long and narrow.

It is possible to map such a subject successfully

using the regular Non-Structures mode flight
planner, but you’re more likely to run into overlap
and coverage issues. Oftentimes, if you use the Non-
Structures mode flight like this, the edges of the
linear plan will have really spotty, uneven coverage
- often ‘stairstepping’ up and down.

Additionally, if the long and narrow subject curves

even moderately, depending on the flight direction,
there might be passes that take the drone outside
of the boundary you’ve created, which creates
inefficiencies in flight and processing. Here’s
an example:

To solve these issues, Business and Enterprise

subscribers have access to the Linear Flight app,
available on our app market. The Linear Flight app
takes only nadir photos, similar to Non-Structures
mode. This means that linear is well-optimized to
produce high-quality 2D outputs, and its 3D
outputs vary depending on the subject.
Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 12
 Here is the same road, except now it is planned with Linear. Notice that
no there are no wasteful passes, unlike above:

The easiest way to plan a Linear

flight is to turn on the Linear
application on a brand new flight
template. Press ‘Yes’ to accept
waypoint modification, and then
‘Save Plan’ to activate the app.

After that, you’ll start out with a

partial box consisting of two green
flight paths surrounding the blue
line guiding them. Move and adjust
the blue line so that it follows the
middle of your linear subject.

We’ve directed the flight to the right line, but only two passes may not provide a wide enough
coverage of the sides of the road.
Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 13
 Especially if you don’t map at a high altitude, there’s a reasonable chance that
the default 2-passes will not be wide enough to capture all of the data for your
scene, or won’t have a sufficient number of images for good reconstruction.
Therefore, it may be important to increase the width of the flight by increasing
the ‘Path Width’ beyond ‘2’. In general, it’s usually better to slightly over-capture
than under-capture.

It is really important to note that processing, exporting, and handling linear

datasets is more difficult than data from more typically-shaped flight plans.
Subpar coverage of an area on a linear map is more likely to cause a processing
problem compared to a more standard map. Please be sure to set altitude,
overlap, and path width as conservatively as possible.

See below for the result of doubling the path width to ‘4’. With this adjustment,
compared to the initial 2 passes, we are now more confident that every inch
of the road will be captured in several images - giving us a little extra
breathing room:

When you’re satisfied with your Linear plan, please be sure to press ‘SAVE
PLAN’ on the app.

Live Map HD
Have you ever flown and wished that you could receive a processed map in real-
time as you fly? Gaining instant insight into the situation on the ground as your
feet are on that very same ground is powerful. Luckily, with Live Map HD, this all
is a reality.

As your drone flies, you can expect to see the drone ‘paint’ the orthomosaic as it
flies on your screen. It can be pretty fun to watch!

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 14
 Live Map uses the live first-person-view (FPV) video
to stitch the map in-real time. This is in contrast to
the full-resolution images from the SD card stitched
together using Map Engine for regular maps.
This means that the Live Map is lower resolution
compared to your full maps, and does not have 3D
data (point clouds, 3D models, and elevation data).
It’s designed to be used for 2D measurements and
scouting with moderate precision.

The default flight settings will work well for the

Live Map. We also recommend avoiding areas that
vary more than 75ft (25m) in elevation. Areas with
small hills and buildings shouldn’t be a problem, but
buildings or terrain taller than 5 stories (50ft or
15m) may cause stitching artifacts in the Live Map.

Because Live Map uses the live data feed from the
drone, losing connection to the drone will cause
gaps or stitching errors in the Live Map, or the
map may fail or stitch incompletely. Connectivity
between the drone and the RC is the main
consideration to take into account when diagnosing
Live Map issues.

Depending on your subscription, you may have

access to Thermal Live Map and Live Plant Health
map, which works similarly to the regular Live Map.

Manual Shots Using Dronedeploy’s

Manual Mode
After your automated flight completes taking its
photos, if there are nooks, crannies, and small
complicated objects that would be difficult to
clearly point out in at least 5 photos, please switch
to DroneDeplopy’s manual mode to provide greater
image coverage of these areas. Further into the
article, we’ll help you easily identify these areas.

Attempting to overlap your manual shots similarly

to how DroneDeploy overlaps the automated shots
is a recipe for success! You will usually only need to
do this in a few areas.

Please be sure to not include the horizon line in

your manual shots, and try to keep the altitude
of the manual shots to within 100 feet of your
autonomous shots.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 15
 Combining Images From Different Flight Modes

Think of each flight mode as a color on your easel.

Depending on your purpose, you can use each
color alone or you can combine them into new
colors. You can also combine image sets taken from
different flight modes into well-balanced maps, such
as adding a Structures mode flight within a Non-
Structures mode flight.

Please make sure that each image section fully

overlaps the surrounding sections, and there are
limitations to combining images with a very large
altitude difference: image sets with more than 100
feet of altitude difference between other image sets
may cause image stitching issues.

After you’ve chosen the flight mode(s) you’ll use for

your flight, now it is time to refine your parameters a
little further.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 16
 03 On the 'Advanced' Panel, you'll find more settings
you can adjust to fine-tune your flight after you’ve
chosen your flight mode.

The Automatic settings by default will be turned on.

This optimizes overlap, speed, and direction of the
‘Advanced’ drone, and will work for most scenes that don’t
have factors making them especially hard to map.
Panel Those factors are outlined in future chapters. If your
subject has any of them, it is likely that you should
turn off Automatic Settings.

Below Automatic Settings, available

to toggle on and off are:
Obstacle Avoidance: We generally recommend
keeping it on, as it can help your drone avoid a crash.
Remember that Obstacle Avoidance is imperfect
and cannot be relied on to prevent all crashes.
Obstacle Avoidance can sometimes cause the
drone to hover when flying directly into the sun
(confusing the sun as an object) or if the lenses are
dirty. Turning Obstacle Avoidance off may enable
the drone to fly faster, and thus capture larger areas
on a single flight plan. Please always manually fly to
check the height of obstacles in the area you plan on
flying to be sure that your drone will avoid them.

Show Existing Map: Planning on an up-to-date base

layer is easier than planning on an old one. ‘Show
Existing Map’ places previous drone maps in the
same area on the ‘Fly’ tab for you to plan the next
flight over.

Low Light: Increases camera ISO to 400, which

makes the camera more sensitive to light, but can
slightly increase graininess or noise in the images.
This is recommended to be used when you’re
mapping in low lighting conditions and the required
shutter speeds are too slow to capture sharp data.

Manual Exposure and Focus: If the automatic

camera (exposure, shutter speed, and ISO) settings
are producing poor-quality imagery, you can toggle
these switches on, and manually adjust those
parameters in the DJI Go app.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 17
 If you turn off Automatic Settings,
you can then adjust the parameters:
Front Overlap: The percent amount that each photo
will overlap with the previous and next photo from
top to bottom, as the drone moves throughout each
pass. Increasing front overlap means that more
photos will be taken one-after-the-other within the
same space. The drone will fly slower with a higher
front lap.

Side Overlap: The percent amount that each photo

will overlap with the previous and next photo from
on each side as the drone moves between passes.
Increasing side overlap increases the number of
passes that the drone will fly over the same area.

With few exceptions, mapping with higher overlap

will increase the quality of your final result
compared to mapping with lower overlap. This
applies until you reach about 90% - 95% overlap.
At that point, the benefits of increasing overlap
diminish rapidly. Below, you can see the difference
between a flight plan with 65% side lap and 85%
side lap:

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 18
 Overlap is a Very Important Setting

Overlap and altitude are two parameters that will most

directly control the quality of your output after you’ve
chosen the correct flight mode. In the next chapter, we will
discuss when to increase your overlap, and in a later
chapter will go in further detail about altitude selection.

Typically, if you’d like to have a 100% fully-stitched

dataset, as you decrease your mapping altitude from 300
- 400 feet, for every 50 - 100 feet of altitude loss, your
overlap must be increased in intervals of 5% to
compensate for the greater difficulty there is in stitching
lower-altitude images.

Flight Direction: The orientation of the passes that the

drone will take. We recommend adjusting this if you are
finding that the drone is hovering repeatedly in the same
spot. We also recommend adjusting the flight direction
parallel to rowed objects such as solar panels or row-crops.

Mapping Flight Speed: This is the maximum speed that

the drone can fly at while it is taking pictures. It won’t
necessarily map at this speed - it will most likely map
much slower than the selected default. We recommend
only slowing down the speed of the drone if the photos
have unavoidable motion blur due to low light, or if the
drone cannot handle image capture requests quickly
enough. Pro-tip: if you’re finding this issue, try purchasing
a faster SD card.

Starting Waypoint: When you move the ‘Starting

Waypoint’ slider, it adjusts the location at which the
drone will start taking photos. If your flight plan was a
16-waypoint plan, and you moved the starting waypoint
to waypoint 10, the drone would fly to the 10th waypoint
to start mapping, and only fly 7 waypoints in total.

Make Available Offline: On the mobile device you plan on

mapping with, toggle ‘Make Available ‘Offline’ to save the
plan locally to your device, as well as saving the satellite
base-map to your device to let you see the background
without internet connectivity. Before going offline, ensure
that the DroneDeploy app is open and logged in. You may
want to turn off automatic updates for your device to avoid
the app automatically quitting due to updates during your
car ride to the field. Please note that apps from the App
Market are online only. After using the flight plan offline,
toggle ‘Make Available Offline’ off to remove the flight plan
and satellite images from your device.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 19
 04
Now that you know which flight tools and parameters you
can use to capture your subject, it is time to think about what
aspects of that subject may present a risk factor for your
successful map. We’ll explain how you can mitigate their
negative effects with the flight tools and parameters we’ve just
Consider How learned about in the previous two chapters.

Your Subject Each factor listed below makes stitching the model more
difficult. Some of them should be responded to differently than
Can Complicate others. But as a good rule of thumb, except when noted, when
you run into any one of these issues, typically you’ll want to
Things And How increase your overlap by at least 5-10% and make sure you’re
increasing your altitude.
You Can Respond
Why Do I Increase Overlap And Altitude To Compensate
For Most Of These Complications?

Increase overlap - Increasing your front and side overlap

increases the number of photos per point, which means that
Map Engine can double-check its work across more images,
increasing our stitching confidence, and thus, increasing
accuracy.

Fly as high as possible - The higher you fly, the larger your field
of view is. The larger your field of view is, the more ‘features’
Map Engine might be able to find per-photo. This also increases
how many tiepoints can be found, which increases stitching
confidence and accuracy.

A Homogenous Subject
The homogeneity of your subject is one of the biggest
factors determining the number of tiepoints we can identify.
A homogenous subject is a subject that looks very similar
throughout each photo. It is hard to pick out distinct locations.
Late-stage row-crops, forests, deserts, metal roofs, and water
are all examples of homogenous subjects.

To illustrate this, imagine in your hand you have two unique

photographs of the same part of a field, such as the image below.
You are asked to point out the same exact same plant between
two photos. This would be very hard for you to do because
there are very few obvious reference points to start from and
everything in the picture looks like everything else. This is the
same problem that Map Engine can run into.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 20
 Most of this image looks the same. It’s ‘homogenous’. It’s difficult to find tiepoints here.

Map Engine has a much easier time successfully stitching together a ‘featured’
map when compared to a ‘homogenous’ one. If you use the default template
settings on a homogenous subject, there is a chance that the map won’t stitch
successfully. Homogenous scenes cause the Map Engine to find fewer tiepoints
than in highly-featured scenes or causes confusion between the same tiepoint
being considered in multiple locations.

When you are mapping a subject with a lot of vegetation, please fly with
increased overlap and fly as high as possible. Increasing your overlap by at
least 10% is ideal, but even more may be necessary.

Elevation Change

A large elevation change in your map or a large difference between the height
of objects in your map can require adjustments to your flight plan. Elevation
change includes tall structures and terrain shifts.

If your overlap values remain constant, as you decrease your effective altitude
over the ground, the density of your image coverage reduces, as seen below:

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 21
“If the terrain of This essentially means that as your ground moves closer
and closer to the drone, the likelihood that there will be
your map changes enough images to stitch a model successfully goes down
as well.
by more than
~20%, you may A 20% elevation increase within your map usually is not
enough to negatively affect coverage enough to make a
need to consider difference. This is true as long as the map is not already near
the fringes of successfully processing.
increasing overlap
To illustrate this, if you flew at 300 feet, and there was a
and increasing 50-foot elevation shift in the area, the section that was 50
your altitude as feet higher than the rest will likely turn out fine. Similarly,
if you flew at 100 feet, a 20-foot elevation increase would
much as possible probably not require adjusting your flight settings for that
20 foot higher section specifically.
to compensate
for the reduced If the terrain of your map changes by more than ~20%, you
may need to consider increasing overlap and increasing
coverage of your altitude as much as possible to compensate for the
reduced coverage of the high area. If the elevation shift
the high area.” is quite large, you may need to carry out the flight using
multiple, different flight templates, in order to keep the
distance from the ground to the drone consistent.

Optimize your Overlap Across Elevation Changes

You can use the Overlap Optimizer app to help calculate

the proper overlap needed given the altitude you plan on
mapping from, and the altitude of your highest obstacle.

To identify the altitude of your highest obstacle, we

recommend manually flying to the height of that obstacle
using DroneDeploy’s Manual Mode.

Intricate/Fine Structures
Oftentimes, you may be interested in reconstructing
objects that are small, thin, and with fine details. This
includes subjects like scaffolding, rebar, external stairs,
statues, or a crane. Successfully modeling these objects can
be paradoxical because you must fly low in order to achieve
the resolution needed to clearly see them in the first place,
but flying low reduces the likelihood of a successful stitch.

This usually means that you’ll need to fly low (<120 feet)
with very high overlap, such as 80-90%. This gives you the
resolution to see the details for these small objects, and
the overlap to compensate for the likelihood of stitching
problems due to low altitude.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 22
 You may want to consider if you need oblique (angled) shots in order to provide
enough vertical and side data for a model. The Precision House 3D app,
crosshatch, and manual flight modes can help produce those kinds of datasets.

Points of Interest and Difficult-to-See Angles

This topic was covered in the Structures Mode ‘Flight Planning’ section, but it is so
important it is worth bringing up again.

Some objects, such as a very simple stockpile, will be adequately modeled from a
basic Non-Structures mode flight. However, many objects will turn out melty and
inaccurate if they aren’t identified as a POI that needs a wide range of side angled
shots, or obliques, in order to be accurately processed.

Even if you add a flight that completely centers around a single object, for some
areas, the autonomous flight will not capture sufficient data. For example, if you
were mapping two tall silos that were very close to each other, it is likely that it
will be impossible to autonomously capture imagery that sufficiently covers the
section of the silos that face each other. Another example might be an alleyway
that is in between two buildings, such as what you see below:

Please take the time to identify each of these areas in the area you are mapping,
then taking a number of manual shots of these areas to supplement the
autonomous ones.

Mapping Rule of Thumb

If you can’t clearly see what you’d like modeled in at least 7+ of the original
images, it is unlikely that it will be modeled well.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 23
 The Shape of the Flight Plan Area
A simple polygon is the best type of shape for flight
planning. The more uniform the flight plan is, the more
uniform and accurate the model is likely to be.

Longer, more narrow linear maps or maps with sharp-angled

hooks, straights, or peninsulas are more likely to run into
coverage issues because there won’t be more than 8 images
of the areas at the tips of the points.

Additionally, it is important to always slightly overfly.

This means to fly a slightly bigger area than you know is
necessary to ensure there’s sufficient overlap at the edges
of the plans.

Oftentimes, the data captured right along the border of

the flight template is insufficient compared to the rest of
the flight plan. You should expect that near the border or
edge of your flight plan, there may be some warping or
misalignment, so if the section you care particularly about is
near the edge of the polygon, please expand the boundary
so that it is at least 100 feet from the edge of the plan.
These problems can be somewhat compensated for with
overlap and altitude adjustments, but the best practice
would be to expand and simplify the flight plan to cover the
larger area evenly.

Reflectivity and Lighting Conditions

We started this article by describing how Map Engine
works. For example, we identify a roof section in photo 1,
and the roof again in photo 2, and then build a model
with the information we can deduce about the roof from
multiple angles.

But what if the roof looks different in every single photo?

This can happen if it is made out of highly reflective
material. If the sun is powerful enough to cause a strong
reflection, an object will have a different reflection in each
location the drone takes a new photo, which makes it very
difficult to match parts of the object to itself.

In addition to the object looking ‘different’ in each photo,

sometimes, a reflection has such a high intensity of light
that it covers any appreciable information about the object
itself as well.

This effect is most commonly seen in the water, metal roofs,

and solar panels.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 24
 To mitigate the effects of a highly reflective surface, we
recommend flying on an overcast day (or at least when the
sun is low), experimenting with camera filters, increasing
overlap, and flying as high as possible.

The sun affects drone mapping in more ways than causing

reflectivity. Flying your drone when the light is flat but still
present will help contribute to the most even and accurate
map possible. Objects are evenly exposed and shadows
aren’t significantly large.

It is still possible to receive great results if mapping while

the sun is more intense, but it may be slightly more difficult
than a more even-tempered day.

Mapping in dusk-like or dark conditions often causes

there to be motion blur and insufficient contrast between
objects on the ground. Poor weather conditions such as fog
or condensation can also contribute to model difficulties. In
these scenarios, you’ll need to pay extra attention to your
camera settings and the quality of your photos as
you fly. You may want to enable ‘Low Light’ or manual
camera settings.

Moving Objects
A moving object captured in your dataset will cause some
degree of inaccuracy. This is because the information Map
Engine receives shows one single object in multiple places
which then has to be placed into one single snapshot of a
moment, which is physically impossible. Moving objects not
only change position but can also change shape, causing
them to be even more difficult to model.

Map Engine does attempt to detect when an object is

moving and to process it in a single place, but it doesn’t
always work. Here is an example of two horses, both of
which were moving while this subject was mapped. One
horse was well-modeled, while the other is inaccurately
reconstructed.

Bodies of water, vegetation rustling in the wind, driving

vehicles, and people walking are the most commonly found
moving objects in maps that will reduce the accuracy of
your map.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 25
 05
You’ve now thought about the flight planning tools at your
disposal and the difficulties that your subject may cause.
You’ve adjusted your flight parameters to compensate for
potential risk factors that your subject may present. It is finally
time to add in other specific project requirements or goals
Tie In Your you or your clients may have, such as relative accuracy, global
accuracy, and resolution.
Accuracy and
Resolution Goals Accuracy
You might have a client or manager say ‘I would like this map
to be accurate to 5cm’. But what does that actually mean in
real terms?

There are two main ways to describe accuracy: relative

accuracy, and global accuracy. It is important to know which
type of accuracy is important for your use case so that your
final result satisfies your project goals.

Relative Accuracy

Relative accuracy describes how accurately the objects within

the map are placed compared to other objects in the same map
area. For relative accuracy, it does not matter how well the
drone data lines up on the Google base layer. The degree of
relative accuracy describes how accurate are the measurements
within the map are to real life.

It is common to require high relative accuracy for measuring

the length, area, and volume of objects, in construction site
management, crop scouting, and creating marketing material.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 26
“High relative
accuracy can
be achieved
without Ground
Control Points
(GCPs) by flying You could have high relative accuracy even if your map was
placed on the completely wrong side of the world because the

low with high measurements within the map would still be correct, but the
geographic coordinates would be incorrect.

overlap and High relative accuracy can be achieved without Ground Control

strong oblique Points (GCPs) by flying low with high overlap and strong oblique
coverage if necessary. With proper GCP technique, relative

coverage if accuracy will often improve a little bit, but primarily, relative
accuracy is a function of the quality of the image set and your

necessary.” drone’s GPS accuracy.

Absolute Accuracy

Absolute accuracy describes how accurately placed your map is

in the world compared to where it truly should be placed. High
absolute accuracy means that if you were to pick a location on
your drone map, and you were to send someone out into the
real world to that same geographic coordinate, they would be
standing at the same location in the real world as you picked on
the drone map.

It is common to desire high absolute accuracy for land title

surveys, as-built surveys, environmental documentation,
and overlaying geo-referenced site plans.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 27
 The accuracy of the location annotation depends on the absolute accuracy of the map.

High relative accuracy is often a byproduct of high absolute accuracy because it

is difficult to accurately place something in the world correctly if its dimensions
are incorrect.

High absolute accuracy can be achieved by properly using GCPs (Ground

Control Points), PPK (Post-Processed Kinematic), or RTK (Real-Time Kinematic)
GPS devices.

When you have instructions to achieve a certain accuracy, we would

recommend making sure it is specified precisely the type of accuracy needed.

How to Increase your Accuracy without using GCPs,

RTK, or PPK
You can achieve a map with high relative accuracy without adding GCPs, PPK, or
RTK. Successfully designing a flight plan for high relative accuracy is dependent
on three main factors:

1. High-Quality Images: Sharp images are paramount for a non-GCP map to

be highly accurate. If an image has blur that goes across a couple of pixels, the
accuracy of the entire map may be halved because half of the pixels are getting
smeared before we process anything. Mapping in good light conditions and
knowing when and how to adjust your camera settings are important for this.

2. High Overlap: We’ve described in this article how more images of the
same location increases how confident we are about an object’s size, shape,
and location. Increasing your overlap will generally increase how relatively
accurate the map will be because we are checking our work across many
images. Very rarely, if there is a lot of movement in your map, high overlap can
have a negative effect on accuracy, because each photo will show something
new and different, instead of confirming what is staying the same.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 28
 3. Fly Low: Assuming risk factors such as a homogenous subject or elevation
change have been factored in, and therefore no part of the model is at risk for
not aligning with the rest of the model at all, and that overlap is properly set,
the lower you fly, the more accurate your map will be. When you fly at 100
feet instead of 400 feet, the images at 100 feet will describe a smaller area
compared to the 400 feet image set but will do so with the exact same number
of pixels. This means that each feature at 100 feet is captured using more pixels
compared to the 400 feet data, and therefore has greater detail. You also have
more images and thus more GPS samples. This allows for the Law of Large
Numbers to have better convergence.

High Precision Mapping

If you’ve flown low with high overlap and good image quality and the relative or
absolute accuracy of your maps is not high enough, using GCPs, RTK, PPK can
roughly increase your absolute accuracy by a factor of 10 compared to a regular
flight. If you add GCPs to your map, you should record your on-the-ground GCP
coordinates by using a high-precision GNSS receiver that is accurate to the
centimeter level.

You Don’t Need GCPs You Do Need GCPs

You do not necessarily need GCPs if If absolute accuracy is important

only relative accuracy is important. then you will need to use GCPs on
Drone mapping projects that only your map. Drone mapping projects
require high relative accuracy that require high absolute accuracy
include: include:

• Measuring length, area, and • Land Title Surveys

volume of objects • As-Built Surveys
• Construction Site Management • Environmental Documentation
• Crop Scouting • Overlaying Geo-Referenced
• Creating Marketing Material Site Plans

GCPs vs Checkpoints?

Ground Control Points and checkpoints are both marked points on the ground
that have a known geographic location. For both, you’ll submit tags, which will
precisely define the center of these markers.

GCPs make your map accurate, while checkpoints allow you to assess how
accurate it is. Checkpoints and GCPs are captured and tagged exactly the same,
except for how they are labeled in the CSV file before uploading them.

GCPs are used as points to ‘snap’ the map to correct locations during
processing. They increase absolute accuracy, but you can’t use their locations
to measure accuracy, because they are algorithmically forced to be near where
their coordinate data says they should go.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 29
 Checkpoints aren’t forced to go in any location. Checkpoints are allowed to
move freely during the processing of the map. Because we know exactly where
they should be, and exactly where they end up, checkpoints are the only true
way to measure the resulting absolute accuracy of a map.

The balance of checkpoints and GCPs you choose will affect how accurate your
map is and how quantifiably confident you can be about the accuracy of your
map. The more GCPs you use, the more accurate your map is likely to be, but
if you have few checkpoints to verify this, it’d be hard to definitively say how
accurate the map is. Using a ratio of 2 GCPs for every 1 checkpoint is a common
starting point, with 20 checkpoints being the ASPRS recommendation for
getting a 95% confidence interval on the results.

We find that Ground Control Point placement is often overlooked. Poor GCP
placement can cause datasets to be more inaccurate with GCPs than without
them! Think of GCPs as having a substantial amount of weight in the accuracy
of the map - this weight, if concentrated in sections of the map without enough
counterweight, can actually cause the map to sag down, tilt, bowl, or even flip.

You should place the GCPs such that if you were to divide out smaller, evenly-
sized sections of the map, they would each roughly have the same number of
GCPs. We recommend planning out the GCP placement before arriving at
the site.

On GCP placement/numbers and map size:

We are often asked how many GCPs are required per acre. Interestingly, one
seldom-considered factor in this equation is the altitude at which you’re flying.
You can think of GCPs as ‘nails’ and each image as a link in a chain. Having a
GCP in an image ‘nails’ that link to the correct place in space. The more links
in the chain between images, the more play there is in the system (the more
the chain will be able to move). Similarly, if we had a GCP in every image, they
would all be nailed down, whereas if there are many images between two
GCPs, there’s more opportunity for sub-pixel errors in the stitching process to
accumulate. We understand that you can’t feasibly put a GCP in every image, so
we recommend 1 GCP for every 60 images.

GCP Placement Rules

• GCPs are evenly distributed throughout the map area

• Minimum of 4 GCPs per map, but 6+ is preferable.
• Roughly 1 GCP for every 60 images (larger maps may scale this down)
• At least 50 feet from the border of the map and from any other GCP
• Use clear markers with a defined center
• Cover the full elevation range of your subject to increase accuracy

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 30
 Here is an example of a good distribution of
checkpoints and GCPs for an 800 image map.
The 13 red stars are GCPs, and the 12 yellow
circles are checkpoints. Notice the evenness
and symmetry of the locations of the GCPs and
checkpoints. There is not an area that is missing
coverage from either the GCPs nor checkpoints.
They aren’t clumped together. With this number
of GCPs, as long as they’re tagged correctly, our
accuracy will be increased substantially over
a non-GCP version, and we’ll be able to verify
that this is the case across the entire map area
by checking the checkpoint accuracies. In a
future section, we will help you interpret this The accuracy of the location annotation depends on the
data correctly. absolute accuracy of the map.

Tagging your GCPs accurately is just as important. Taking the time to precisely click the middle
of the target and to verify that there are no stray, erroneous tags that are off-center or of a
different GCP is the final ingredient leading you towards the next level of accuracy. Please tag
at least 6 views per GCP.

Notice how each GCPs’ tags are centered and look similarly exposed
and focused.

Here is an example of poor tagging. 4 out of the 6 tags are off-center. This will create a large
degree of inaccuracy for at least the area around the GCP, if not, the entire map.

Notice how each GCPs’ tags are off-center compared to the first image

You can also increase global and relative accuracy by using a drone with RTK or PPK. When
you fly a normal drone without RTK or PPK technology, its GPS accuracy will be similar to your
smartphone - typically within a few feet, which is not very good for high-precision mapping,
and is the reason why non-GCP maps have similar accuracy limitations.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 31
 PPK and RTK systems mean that the drone can record
its own location very accurately, which means that the
information we’re receiving to build our model is more
accurate. Oftentimes, the relative and global accuracy of
PPK and RTK maps is good enough such that no GCPs are
needed to increase accuracy, this can be a time-saver and
simplify the workflow.

There are still reasons to use some GCPs if you’re RTK

mapping, for example, only checkpoints are able to verify
that the drone data is accurate.

“PPK and RTK Interpret Your Accuracy Report Data

Correctly
systems mean
that the drone
You now know the difference between relative and
absolute accuracy and if they’re important to your project

can record its

goals. You know how regular flight techniques, GCPs, and
checkpoints can impact absolute and relative accuracy.

own location
Now, to conclude our section about accuracy, we’re going to
explain the accuracy report data. What does it mean,

very accurately”
and what can you infer from it?

When you map (1) without GCPs, (2) with only GCPs, (3)
with checkpoints, the accuracy data you receive in each of
these three scenarios is a different and unique metric. So, if
you want to tie this data back to those relative or absolute
accuracy goals, it is important to know what data you are
actually receiving in the report and what you can and can’t
deduce from it.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 32
 The glossary in the Accuracy Report for your maps contain these specific
definitions. Here is a summary of how you can use your Accuracy Report data

Map Type Data Type Received How can it be used?

No GCPs or The camera location XYZ The camera location XYZ

checkpoints used root means squared error RMSE cannot be used
(RMSE) is the average image to determine relative or
location error. absolute accuracy.
It can sometimes give you a
It’s saying ‘How close were general estimation for the
the drone images’ geotags to relative accuracy of your map.
the locations where we found
the images needed to be in For example, if your camera
order to make a model that location RMSE was less than a
makes sense?’. few meters, it’s likely your map
has solid relative accuracy -
This is more of an indication but not a guarantee.
of the accuracy of a particular
processing step of the Map If your camera location RMSE
Engine, and less an indication is above 10 meters, we would
for the accuracy of the entire recommend taking a closer
map. look at the map. It, however,
still could possibly be accurate.

Just GCPs used The ground control point The GCP XYZ RMSE cannot
(GCP) XYZ root mean squared be used to determine relative
error (RMSE) is the average or absolute accuracy.
GCP location error in the
XYZ axis across all the These values are calculated
processed GCPs. using a weighted
mathematical estimation,
It’s saying ‘How close was because we ‘force’ the GCPs
the GCP location as measured to go very close to their CSV
by your precision GPS device coordinate locations. When
compared to the corrected we ‘force’ the GCPs to move
GCP location that is calculated into a particular location in
during map processing?’ order to help rectify the rest
This is more of an indication of the map, the RMSE values
of the accuracy of a particular become algorithmically
processing step of the based, instead of based on the
Map Engine, and less an accuracy of the map.
indication for the accuracy
of the entire map.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 33
 ...Continued

Map Type Data Type Received How can it be used?

Checkpoints used The checkpoint XYZ root Checkpoint location error

(with or without means squared error (RMSE) is a measure of the absolute
GCPs) is the average checkpoint accuracy of your map. This
location error in the XYZ metric seems similar to
axis. It’s saying ‘What is the GCP XYZ RMSE. The
the difference between difference between the
the checkpoint location as two is that checkpoints
measured by your precision aren’t weighted. They aren’t
GPS device and the correction ‘forced’ to go anywhere. That
checkpoint location that way, they ‘naturally’ move
is calculated during map completely with the rest of
processing.’ the map. Remember that the
location error for a checkpoint
only describes the locational
accuracy of that single point.
It doesn’t actually describe
the locational accuracy of
the entire map, because it
is normal to have different
accuracy in different parts
of the map. So, the more
checkpoints you use over a
wider area of your map, the
better the estimation they’ll
provide for absolute accuracy.

So, how can you check for relative accuracy? You might have noticed that in
the above chart, none of the data points are a true representation of relative
accuracy, although if the absolute locations of the checkpoints are accurate, to a
degree, the distances between them will also be accurate.

Remember that relative accuracy describes how accurate measurements within

the map. So, if we know the dimensions of some objects on the ground, we can
use ‘ground-truth’ comparisons to establish a relative accuracy estimation.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 34
 For example, when you’re capturing your drone data, you might take a tape
measure and take down the dimensions of clearly defined, static objects
distributed throughout your map, such as the objects circled below:

Then, measure those exact same locations on the completed drone map, and be
sure to be as precise as possible. The average percentage difference between
the actual and the reported measurements can give you a good idea of relative
accuracy. Here is some sample data:

Feature Ground Drone Map % Difference

Measurement Measurement

White sidewalk 7.25 meters 7.50 meters 3.4%

section

Reflective parking 6.65 meters 6.74 meters 1.3%

lot strip

Bridge 6.30 meters 6.18 1.9%

The average of the three percent differences is 2.2%. Therefore, the area of the
map that these objects cover is likely to have a near 97.8% relative accuracy,
and that would be a good starting point for the accuracy of the measurements
you make in this area of the map. Just like checkpoints, bigger diversity objects
measured over a wider area of your map will help estimate accuracy better.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 35
 Resolution
As you change mapping elevation, what changes is the size of the ground area
in which each pixel covers. This is called the Ground Sampling Distance (GSD),
and this is usually the metric someone is looking for when they say something
along the lines of ‘I would like the resolution to be 2cm’; they usually mean
2cm per pixel.

The lower you fly, the smaller your GSD will be, which means that each pixel,
when mapped on the ground, will cover a smaller area. This means the map
will be clearer and less pixelated than the maps you fly at a higher altitude.
Remember to compensate with overlap increases as you decrease your
mapping altitute.

Selecting Appropriate Altitude

Each image from the same camera is a fixed number of pixels in size. If that
camera is close to an object, then many of those pixels will be of that object. If
the camera is far away from the same object (and at the same zoom level), then
there will be fewer pixels of that object. This is the principal behind GSD. As
the drone is closer to the ground, each image will capture a smaller section of
the ground, but it will have many more pixels for the same surface area of the
ground compared to an image captured from a higher altitude. This increases
the accuracy of the map.

As you change the altitude in your flight template, the expected resolution of
your map will also change. Remember that until your drone is connected to the
DroneDeploy app, we won’t be able to know which camera (or lens) you are
using, and so we will be estimating this based on the most popular drone used
for mapping - the Phantom 4 Pro.

As you reduce your altitude in flight planning, you’ll also see an increase in the estimated resolution.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 36
 06
If you’ve made it this far, you know that there are a lot of factors
to consider when designing a top-notch drone flight plan.
Depending on your subject and goals, there may be factors
that produce opposing advice or seem to be in conflict with
each other. There may be a few factors that push you to fly low
Create and others that encourage you to fly high. But, now you know
what flight planning tools you have at your disposal to create a
A Flight Plan balance between all of these forces.

You can balance these forces, such as desired accuracy,

homogeneity, or the intricacies of your subject by first,
compromise. For example, if you have reasons to both fly
high and fly low, somewhere in the middle will likely be the
best choice.

Secondly, balance these forces by compensating. You can

counteract a shortcoming that one setting causes by changing
another. In the same example as before, if you chose a moderate
altitude when a higher one would be more ideal, you could react
to that further by increasing overlap.

Lastly, redundancy and experimentation really pay off. While

you’re still mastering drone mapping, create alternative flight
plans using different techniques and settings and try processing
different versions of your subject.

For example, you might process one map with manual shots
added and one without - see if the difference is enough to
matter to you. You might also compare a Structures flight vs.
a Precision 3D app flight of the same exact subject. Or, try
to process one version in Structures mode, and one in
Terrain mode.

The more maps you create using a wide range of settings, the
faster you’ll learn how each decision you make in the planning
phase affects the final map.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 37
 07
Using the officially supported mobile devices and drones are
important for safety and performance reasons. Before every
product release, we make sure to test every drone, camera, and
lens combination listed in our supported drones list.

Bring Your Be warned that unlisted devices are not frequently tested by
our team, and we are unable to guarantee support for those
‘A’ Game To platforms.

The Field!

A few of the most commonly used drones on DroneDeploy.

It’s also important to have a consistent routine for each mission

you fly. Here is a list of some common workflows that will give
you the best chance of having hundreds of successful flights.
This is not the ‘be-all-end-all’ list, but a great starting point to
make sure you’re doing your due diligence:

1. Check the iTunes Store or the Google Play store for updates
to DroneDeploy. We recommend turning on automatic updates.

2. Use DJI Go or DJI Assistant to upgrade to the latest firmware

for your drone and remote controller, and make sure your
aircraft is fully calibrated (especially the compass). Uncalibrated
hardware and out-of-date firmware are the two most common
issues that prevent people from flying or causes unexpected
behavior.

3. Inspect your drone for damage before going out in the field to
ensure that it’s ready and safe to fly.

4. Fully charge your remote controller, mobile device, and

drone batteries.

5. Format your SD card before going out to fly to make sure you
have sufficient space. Always have a spare SD card on hand in
case the card in the drone gets corrupted (or in case you leave
the drone’s SD card in your computer!)

6. Bring 2 USB cables for your mobile device, just in case

one malfunctions.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 38
 7. Bringing additional propellers for your drone can help ensure you can keep
flying even if propellers get damaged during your flight.

8. Utilize websites like weather.com to ensure that the area you will be flying
in has good weather.

9. Scope the site out using Google Earth, and Google Streetview ahead of time to
get more familiar with the site.

10. To ensure that the area you will be flying is cleared for drone use DJI Fly Safe.

11. Familiarize yourself with applicable local, state, and federal regulations.

Choose your home point and the location of your controller strategically. You
should always maintain a line of sight with the drone, and you want to promote
the strongest signal between your drone and your controller as possible. You also
don’t want to have significant altitude changes in your starting location between
battery changes.

To do this, if possible, take off from a high point of the map, as well-centered
within the area as well, and do not plan a flight that has a larger radius than the
signal range of your drone. Avoid taking off next to sources of interference, such as
physical barriers like big buildings, hills, and trees, and electromagnetic interference
sources such as power lines, reinforced concrete, and large metal objects.

You must check the altitude of the notable obstacles within the entire area that
your drone will fly. To do this, use Manual Mode in DroneDeploy to manually fly
until the horizon is above all nearby obstacles to ensure your drone will fly over
them with a comfortable buffer. Obstacle Avoidance can help avoid obstacles but
it is not perfect.

As you fly, always be in command of your aircraft, fly within line-of-sight, and
always be ready to take over manual control of a drone if something arises.

Check your Image Quality as you Fly

DroneDeploy does attempt to make your camera capture imagery at its absolute
best quality. However, ultimately image quality is governed by so many factors that
it is useful to check the quality of the individual photos captured as the drone is
performing the flight. Check that your subject is evenly exposed and in focus. The
easiest way to do this is to tap the FPV thumbnail so that you can view the FPV
full-screen.

You can toggle the “Set Exposure/Focus Manually in DJI Go” setting under the
‘Advanced’ tab while planning your flight toggles to adjust your exposure and focus
manually in the DJI Go App. You may need to restart the flight after adjusting your
camera settings if the automatic settings are not producing great results.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 39
 08
As soon as possible, we recommend inspecting your images on
a desktop or laptop computer. If your images have problems,
the sooner you find out, the sooner you can refly or reassess.

Your images should be evenly exposed and sharp. You should be

Process Quality able to zoom in on them a fair bit before they get soft.

Data in the Notice how you can see the details in the roof when zoomed in,
even though the picture is from fairly high up. Both the white
Correct Mode sections and the grey sections are properly exposed, which is
also important.

In order to use the Map Engine, your data must meet the
following requirements:

1. All images should be in JPG format

2. All images must have latitude, longitude, and altitude in the

GPS EXIF data

3. All images should be facing the area of interest

4. All images should have significant overlap (more than 70%

side and front overlap)

5. At least 30 images for reliable map processing.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 40
 Here are some common image problems to watch out for. These will all cause
Map Engine to find fewer tie points, and will decrease the quality of your model.
Remember, outputs (maps) can only be as good as the inputs (images).

Motion blur, usually caused by a Soft or out of focus, usually caused

slow shutter speed or low lighting by incorrect aperture setting:
conditions:

Overexposed, usually caused by slow Underexposed, usually caused by fast

shutter speed, high ISO or too large shutter speed, low ISO or too small of
of an aperture: an aperture:

Horizon included, caused by too high

of a camera angle:

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 41
 Processing Modes
We have 2 main processing modes that you can choose to
process your data in. The first is the Terrain mode, and the
second is Structures mode.

Non-structures mode flights typically receive Terrain Mode

processing, and Structures mode flights typically receive
Structures mode processing, but there are reasons that you
may choose to process your data differently. It can often be
crucial that you process in the right mode for your subject
and project goals. Note that Terrain mode still generates
3D data products like Point Clouds, 3D Models,
and Elevation Layers.

You can choose which mode to process in as you upload

your data.

Terrain mode is optimized to:

• Reconstruct large tracts of land such as agricultural

fields or site survey accurately in 2D
• Reconstruct 3D surfaces that do NOT contain
overhangs or vertically stacked points.
• Process datasets up to 10,000 images.
• Process nadir-only datasets.
• Provide a gridded raw point cloud that is 4x less dense
than the Structures mode point cloud.
• Fill in holes more aggressively in Structures mode.

Structures mode is optimized to:

• Reconstruct 3D objects as accurately as possible.

• Process oblique imagery along with nadir shots.
• Process image sets of up to 1,000 images. If your
dataset is a little bigger than that, you can still try to
process it.
• Provide a point cloud that is 4x more dense as the
Terrain point cloud, with data in overhanging regions
preserved, and the points filtered less aggressively.

You can bend the rules strategically if the results you

receive the first time around aren’t what you would
have hoped for.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 42
 For example, if you have Non-Structures-style data (nadir
only), but you were hoping for a better 3D result than what
you received when you process the data in Terrain mode,
you could reupload the data and process in Structures
mode to try for a better result. The dataset is not designed
for Enhanced Structures, but experimentation can lead to
positive results sometimes!

Likewise, if you had Structures style data (including some

obliques) that had coverage issues and had a few holes, you
can certainly try to reupload the dataset in Terrain mode.
You would see a reduction in 3D quality, but it may fill the
holes more effectively.

Any dataset that includes obliques up to 1,000 images

needs to be processed in Structures mode if you do hope for
those obliques to be used effectively in processing to help
the 3D model.

The processing time of your map varies highly depending on

the size and quality of the dataset.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 43
 09
Once your map completes processing, there are
steps you can take each and every time to help
you understand how accurate your data is and the
reasons behind its accuracy. These steps are the
basic troubleshooting steps that DroneDeploy
Quality support uses when you reach out with a question
regarding the quality of your map.
Check Your
The first step that we recommend taking is a sanity
Data and Map check using your eyes. We have a lot of tools and
data points that can help you determine or infer the
accuracy of your map but don’t discount the power
of your own pattern recognition.

Switch between the 2D and 3D views. In the

3D view, move around and double-check that
everything looks similar to how it looks in real life.
Where you start to see reconstruction problems,
you are more likely to run into measurement
inaccuracies. If part of the map is processed
incorrectly, you are very likely to catch it over this
step.

In the 2D view, check the alignment of the map

against the base layer. Remember that without
GCPs, your map may be at least a few meters off of
its precise location because drone GPS data is not
very accurate. Do also keep in mind that the Google
base layer can be imperfect as well.

In this example, the top two circled areas are very

well-aligned with the base layer. There is probably
solid absolute accuracy in those sections. The lower
circled area shows some warping near the bottom
because this is near the edge of the map, where the
image coverage is light. This warping indicates that
this area is not accurately reconstructed, and so we
would recommend avoiding trusting measurements
in areas like this.

Now that you’ve inspected the map and understand

its strong and weak points, relating those strengths
and weaknesses to your image set will really begin
to shape your understanding of photogrammetry
and your ability to strategically plan your flights for
beautiful results.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 44
 The second step we recommend taking is to turn on the image projections. Each
and every time you process a map, we recommend going to the ‘Media’ section
of your map and turning them on.

Consider the above picture. Why did that section turn out misaligned?
Because the flight plan boundary was near the area marked in green below,
and oftentimes, there is insufficient image coverage around the edges of the
map, especially around a corner.

Oftentimes, poorly reconstructed areas have poor image coverage, and so if an

area did not turn out as well as you would have hoped for, check to see if there
is a continuous, dense block of photos directly over the problem area is a great
step to take. Preferably, image projections might remind you of a thick blanket.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 45
 Your image set should not be slightly overlapping, disjointed, or scattered.
You can see how this single bridge of ~20% overlapped images in thick
homogeneous vegetation is might not provide enough data to construct
a model.

Not only can you directly see the

placement of the images in the user
interface, but you can also inspect a
heat map that shows the number of
images per pixel.

Areas in blue have good coverage,

areas in green have medium
coverage, and yellow and red areas
have poor coverage. This can very
quickly help you diagnose coverage
issues such as holes or warping.

You can see how the poorly-

reconstructed area also has
little to no images per pixel:

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 46
 However, it is very possible to have what visually looks like dense coverage of
your subject but is in fact not. It is a classic ‘quality vs. quantity’ situation. Going
back to our ‘Flight Planning’ section, if there are not enough oblique images that
cover the sides of your objects, 10+ images covering each pixel in that area, then
there will not be enough data for an accurate reconstruction.

If a building looks melty or poorly reconstructed, drop a location pin on it and

move it around. Inspect the images that cover it. If you can’t very clearly point
out what you would an accurate model look like in at least 5 photos, it is likely
that better coverage is needed.

Next, take a look at the accuracy numbers for the report. The “Tie in your
Accuracy and Resolution Goals” outlines what you can extrapolate from the
RMSE accuracy data of your map, depending on your usage of GCPs and
checkpoints, and how these numbers are calculated.

Remember that most of the time, this data can only be used as a signal. You
can think of the RMSE accuracy number loosely as a risk factor. The lower the
number, the more likely that your map is accurate.

It is difficult to pick a particular RMSE value that you should begin to be

concerned with. The best way to decide your threshold for RMSE accuracy is to
look at the accuracy of similar maps you’ve processed in the past. If you were
happy with the accuracy of those maps, the average of their corresponding
RMSE values might represent a ‘green zone’ for you. If you were unhappy with
the accuracy of maps in the past, those RMSE values might be a ‘red zone’, which
may trigger you investigating the map more closely.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 47
 If you used GCPs or checkpoints, it is important
to check out the accuracy report. Most GCP maps
should have RMSE errors below a 1 meter. If your
accuracy is worse than that, inspect the accuracy of
each individual GCP and checkpoint.

Look for GCPs and checkpoints with outlier

accuracy values. Oftentimes, if your GCP maps’
overall accuracy is poor, there are outlier GCPs that
have considerably worse accuracy than the others.

Those outlier GCPs are often mistagged, so please

go back to your GCP tagging link to inspect their
tags. You may want to retag your problem GCPs and
reprocess. There also may be something wrong with
the coordinate data or the image coverage of the
GCP. If the GCP or checkpoint cannot be improved
to a level closer to its peers, consider throwing it out
of processing, as long as it is not in an essential part
of the map.

With these final ‘Quality Check’ steps, you finally

can be qualitatively and quantitatively sure that
you’ve successfully mapped, and can repeatedly do
so throughout a wide variety of conditions.

Happy Mapping!

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 48
 10
To end our guide, we’re providing you with easy,
bullet-pointed recipes for the basic direction to
take when mapping for a specific project goal.

Recipe Book
for Common
Project Goals

Important Considerations For 2D Layers:

1. Fly as high as possible to increase the percentage

chance of successful stitching.

2. Take nadir shots for the highest possible accuracy.

3. Use a Non-Structures mode flight.

4. Maintain a drone-to-ground elevation change of

no more than 50-100 feet.

Important Plant Health Map Considerations:

1. Increase overlap due to homogeneity.

2. Fly as high as possible in Non-Structures mode.

3. High resolution and accuracy are usually

not needed.

Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 49
 Important 3D model considerations:

1. Structures mode, or Precision 3D app to ensure

orbits for side coverage.

2. Fly low for greater detail.

3. Extremely high overlap because of low altitude.

4. Structures requiring the highest 3D quality

possible should be centered as a point of interest
in a Structures mode or Precision 3D flight.

Most important Elevation layer considerations:

1. All 2D layer considerations listed in the first

recipe apply here.

2. Sometimes more coverage of different angles and

elevations is needed compared to just a 2D layer.

3. Flying lower increases elevation accuracy, but

at the expense of the percentage chance of
stitching success.

4. If you fly low you need to then increase the

overlap to compensate.
Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 50
Making Great Maps: The Complete Guide to Professional Mapping with DroneDeploy 51