Post-Processing for drone mapping

——MAPT

Kristof Van Oost

Emilien Aldana Jague
He Zhang

25/4/2018
Sections: 1. Drone-mapping workflow
2. RTK and PPK
3. How to use RTKlib
4. Cloud and DSM generation in Pix4D
 Drone Mapping Workflow

Process
Capture

Analyze
 Tie Points and Overlap

CREATING A MOSAIC BY STITCHING SEVERAL ORTHORECTIFIED IMAGES.

SOURCE: (MEO ET AL, 2012).

Figure 1. Using a tie point to represent a feature

A tie point (TP) is a feature that you can clearly identify in two
or more images and that you can select as a reference point,
as shown Figure 1.

The perspective center coordinate provides a strong geometric

Figure 2. Tie points in overlapping images
tie between the models, helping recognize adjacent images.
 RTK and PPK

Non-RTK drone

-with an onboard GPS receiver

-Normally, you will encounter an inaccuracy up to several

meters, especially in Z directions.

In that case, we need ground control

point (GCP) to help improve
positional accuracy in processing.
 RTK and PPK

RTK drone

-Connect to a base station or a virtual reference station (VRS)

-Camera positions are calculated in real-time, up to 2 to 3 centimetres

of accuracy

Post-Processing (PPK)
-In some cases, direct connection is not available
-We can perform the computation referring the ground
reference afterward with a PPK drone system
 Setup
EUREF Permanent
GNSS Network

Corrections
From Base Station

Drone with Reach Kit

Figure. Distribution of EUREF in Belgium

•Base stations is a Reach unit in Wi-Fi AP mode, configured

as NTRIP. Mount Point:BRUX00BEL0 Reach RS
Data management
 RTKlib

For Post-Processing, we use RTKlib

For visualization

For converting UBX to

OBS and NAV file

For post-processing
 RTKlib
Step 1. Start RTKLIB RTKCONV after downloading raw files
from Reach to your PC.
 RTKlib

Step 2-1. Start RTKLIB RTKPOST

from Rover

from Base

from Rover

First configure the options

 RTKlib

Step 2-2. Set Base Position Open the LLH. File

with RTK_Plot, find the Fix point.

Single

Float

Fix

We can get it from “Solution file”

in RTKPlot
 RTKlib

Step 2-3. Set Setting1 and Setting2

Personal experience: Due to signal quality, if you use different settings, you may
acquire different positional output.
 RTKlib

Elevation mask Integer Ambiguity Res

 RTKlib

Step 3. Check the

position reliability using
RTKPLOT
A shift between “fix and
hold” and “continuous”

[False Fix]

After processing, we got POS file and event.POS file which can be open with RTKPlot.

We compare the positions from different settings, if they are all close (within centimeter),
meaning that the position is trustworthy. In this case, we use “Fix and Hold”.

Otherwise, if the positions vary from different settings (meter level), “Continuous” is more
trustworthy.
 RTKlib
Step 4. Arrange Coordinate Info in NotePad++ and Excel

Open events_POS file with NotePad++, check the number of recordings (whether match photo numbers).

Note that: the first recording may be the “switch on” of camera. Distinguish by time interval.

Q: quality
1:fix
2:float
5:single
 RTKlib
Step 4. Arrange Coordinate Info in NotePad++ and Excel

~22cm
Remember to remove the antenna height of the drone.

Save as CSV file

 RTKlib

More information about Reach manual, go to:

https://docs.emlid.com/reach/common/tutorials/gps-post-processing/

More information about RTKlib, there is a manual:

 Geotag the photos in R (optional)
Put photos and
coordinate.csv
and exiftool.exe
in a folder.

Write GPS info

into JPG files.
 Processing in Pix4D
Now we have geotagged Photos, we could process in Pix4D or Agisoft Photoscan.

Step 1. start Pix4D, create new project

Next page
Processing in Pix4D
 Processing in Pix4D

If we haven’t geotagged the

photos, we need to import
the coordinate from file.

It’s the liberty you give to the

software, that to what extent it
can trust the coordinate of photos.

There is an positional adjustment

through calculation.

We can set 0.05 m here.

Next page
 Processing in Pix4D

Next page
 Processing in Pix4D

Here we use
3D Maps
 Processing in Pix4D
Step 2. Select images and GCPs

Select images
(remove images:
during taking off and
landing, flurry, turning)

Adding GCPs.
Here we use GCP only
as Check Point.
Note: choose a proper Geo
System
 Processing in Pix4D
Step 2. Select images and GCPs
GCP or CP ?
For images For GCPs Here we use CP

Remember
to set a
proper
Coordinate
System

Uncheck the images

you don’t use
 Processing in Pix4D
Step 3. Configure Processing Options

uncheck
Uncheck (for saving time)
 Processing in Pix4D

uncheck
 Processing in Pix4D
Step 4. Run Initial Processing

Now we can run the

processing:
1. Initial Processing.

After that, we will have

a rayCloud view.

Then we can select

GCPs, mark the
location in the images.

Note that: this step is for calibration with GCP or

validation with CP. If we are going to use direct
georeferencing (without GCP survey), we can
skip this and directly run all the 3 steps.
 Processing in Pix4D
Step 5. Marking GCP and reoptimize

As for marking on the images,

we select the GCP and click its
position on the images,

After several clicks, we could

do it by automatic Marks. We
make sure each mark is
correct, then we click Apply.

When finish the marking, we

do “reoptimize” in “Process”
menu.
 Processing in Pix4D
Step 6. Run Following Processing

Now we can run: 2. Point

Cloud and Mesh and 3.
DSM, Orthomosaic and
Index

After that, we will have

generated Cloud and
DSM.
 Processing in Pix4D

Now we can do
what we want.

Eg.
Cloudcompare
ArcGIS

Figure. Topography change between two survey

 (a)

Before plowing
(b)

Reference cloud

After plowing

Figure. Outputs of cloud-to-cloud (M3C2) of reference cloud and (a) cloud before plowing (b) cloud after plowing