IN THIS ARTICLE
Where is this functionality implemented in the software?
How much faster is the accurate pipeline?
How accurate should my IMU be?
As a user/drone manufacturer, what do I need in order to use the pipeline?
What is the type of angles I need to provide? Roll, Pitch and Yaw or Omega, Phi, Kappa?
How could I check whether my drone correctly tags the Exif data?
What are the default IMU accuracy values?
What happens if my IMU accuracy values are too confident or too pesimistic with respect to real values (e.g. my roll angle is 2 deg accuracte but there is 0.1 deg or 10 deg in the Exif)??
Do I need to have an RTK drone in order to use this pipeline?
Where is this functionality implemented in the software?
- On the menu bar click on Process > Processing Options…
- Select step 1.Initial Processing
- Select the tab Calibration
- In the section Calibration, select Accurate Geolocation and Orientation as calibration method.
For more information, please refer to this article.
How much faster is the accurate pipeline?
On average, the speed increases by 20%. On larger projects, the increase in speed is up to 40%.
How accurate should my IMU be?
- Given an IMU accuracy of 3 degrees or lower, the processing-time decreases noticeably.
- The use of precise IMU does not improve the accuracy on the ground, e.g. at checkpoints.
As a user/drone manufacturer, what do I need in order to use the pipeline?
You need to have the RTK/PPK geotags inherent to your XMP Exif Tags.
What is the type of angles I need to provide? Roll, Pitch, Yaw or Omega, Phi, Kappa?
- We expect Roll, Pitch, Yaw angles.
- Pix4Dmapper is capable of accounting for the residual angular misalignment (AKA boresight) provided it is constant (does not change between images) and is small (<5 degrees).
More information on how Yaw, Pitch and Roll are defined can be found here.
How could I check whether my drone correctly tags the Exif data?
Create a project and open the .p4d file and search for the following tags:
<image>
........
<accuracyXY>...</accuracyXY>
<accuracyZ>...</accuracyZ>
<accuracyOri yaw="..." pitch="..." roll="..."/>
.........
</image>
The values you see should match what you expect (e.g. with an RTK/PPK you may expect 10 cm accuracy, so you should see 0.1 in accuracy XY). The same applies for the IMU accuracy.
Note that the accuracy values represent the sigma of the normal distribution (67% of the values are within +/- 1-sigma and 99.5% fall in +/- 3-sigma), so check with the manufacturer to know what is the accuracy they sell you.
What are the default IMU accuracy values?
In absense of IMU accuracy data in the Exif data (example: DJI Phantom 4 Pro), Pix4D sets default values of 10, 10, 15 deg (Roll,Pitch,Yaw).
What happens if my IMU accuracy values are too confident or too pessimistic with respect to real values (e.g. my roll angle is 2 deg accurate but there is 0.1 deg or 10 deg in the Exif)?
- The accuracy values need to be realistic.
- Non-realistic standard deviations of the angles may negatively affect the processing quality in all calibration pipelines! Although the effect on the standard pipeline is negligible, over confident IMU accuracy tags may affect the processing.
Do I need to have an RTK/PPK drone in order to use this pipeline?
Ideally yes, but it is not necessary in all cases. We have seen that this pipeline helps even with less accurate drones, thermal projects...
Hello. I used Sequoia multi-spectral sensor with Sunshine sensor (which has IMU incorporated) for my project. I wonder if the accuracy level of the images is sufficient to apply "Acurate Geolocation and Orientation" in the Calibration (step 1).
Could anyone advice me?
Thank you.
Hi Giorgi,
Accurate Geolocation and Orientation is recommended for the project with very accurate image geolocation and orientation. So, I wouldn't recommend using it for the sequoia camera. In the case of the multispectral sensor, I would highly recommend you to use Alternative Calibration because it is best suited for the low texture content and relatively flat terrain, fields for example.
Hello
My project is 6000 photos (Nadir), 1100 HA area, and almost flat terrain.
Why can't I use Accurate Geolocation and Orientation in my project?
What can I do to use it?
Hi,
The Accurate Geolocation and Orientation calibration method is intended for images that have very precise geolocation information (e.g. RTK or PPK drones). Usually, the geotags provided by these surveyor grade drones contain image geolocation with an accuracy of few centimeters.
The option can only be selected when you have both coordinates and angles (omega, phi, kappa) for each image and you are confident that they are precise. If these are not accurate, the calibration will fail in the majority of cases.
Cheers,
Teodora
Hi,
would the Accurate Geolocation and Orientation calibration method be a good option for the Phantom 4 RTK when the External Parameters Optimization is set to Orientation?
Values from .p4d- file below.
Hi,
If you choose the calibration method Accurate Geolocation and Orientation, the External Parameters Optimization is automatically disabled (i.e. set to None). We recommend this type of calibration if your image geotags are highly accurate (RTK/PPK), which means that the camera orientation and position are already very precise.
Cheers,
Teodora
Hi,
The accuracy we are getting from the RTK is under 20mm, even so, when calibrating the images in Pix4Dmapper using Accurate Geolocation and Orientation method, we are still seeing a slightly shift of about 1m.
We used a Trimble R10 ground station to record some GCPs and to check the Outputs accuracy against them.
Unless we tie up the data to the GCPs coordinates, the 20mm accuracy from the RTK looks useless.
Any thoughts?
Thank you muchly!
Hi Dan,
Here are some factors that you need to consider:
1. Make sure the RTK correction data has been correctly imported in Pix4Dmapper.
2. If you really want to know the accuracy of your project, you should use check points, which are not included in the georeferencing process. The fact that the RMS error values associated with the GCPs are showing very large values does not reflect the exact accuracy. Please use 2-3 check points and evaluate their accuracy once more.
3. Your project might have areas where the keypoint matches are not so good (i.e. an indicator for few connections). If there is a GCP in that area, it will not be so precise.
4. The overall accuracy also depends on the accuracy of the GCPs themselves. For example, if you mark the GCPs visually, you can have a larger error in the model itself. Please ensure that the GCPs are homogeneously distributed across the dataset (to avoid any georeferencing bias), marked at a high zoom level and in as many images as possible.
I recommend also sharing this issue with other savvy users on our community forum, they might provide more hints which will help you troubleshoot faster: https://community.pix4d.com/c/pix4dmapper/
Best,
Teodora
Hello Teodora,
Could we use Phantom 4 Pro with PPK system from Topodrone or other supplier with this pipeline? When importing the images, there were values for Omega, Phi, Kappa.
Hi Veha,
If you have the orientation values, you can use this calibration pipeline. Since you are using PPK, we assume the accuracy will be high enough. Also, use check points to access the accuracy of the project in general.
I have a workflow where I produce multiple subprojects which I process using the "accurate location and orientation" pipeline.
This goes well.
I then merge the projects.
This goes well.
I then go to select my processing options and have the warning:
"Invalid value in calibration tab".
When I change this value to standard, processing works fine.
Nevertheless, this is frustrating as I have to reoptimise before continuing, which takes time.
Why is this happening? All of my images have accurate geolocation and orientation and processing completed just fine prior to merging.
Hello, similarly to your previous question about merging, we need to have a look at the quality reports of the individual projects (before merging) and the quality report of the merged project (before the Reoptimization). We need to check if all of the individual projects have been processed with the accurate geolocation pipeline. Please create a new community topic here attaching the reports and a screenshot of the warning you get. Thank you.