Support Website Contact Support Blog

Poor Mapping Result With GCPs

I am fairly new to Pix4D but have combed through the available videos and articles to gain as much of an understanding as possible. I have been using other photogrammetry software for some time, and found the quantitative results in P4D attractive. During my first project while flying a Phantom 4 RTK we ran into some issues in the field. We were using the D-RTK 2 pole, but were unable to attain RTK lock, or even a floating RTK connection. Deciding that PPK would be appropriate, we carried on with our manual flights.

I flew the same site at various elevations, nadir, oblique and some strings of images that came in close to the control points, and then panned out. Having run the images through P4D, adding MCP and GCP, I have a truly awful result. I would sincerely appreciate any assistance in resolving this, as the only other solution is to return to the site and fly it again, hoping for better results. For obvious reasons, this is the last resort.

I am not terribly knowledgeable on Geodetic surfaces and datums, nor will I understand any surveyor-type lingo. Any help is much appreciated! Picture attached for reference.

Hi,

Thank you for reporting the issue.

Could you share with us your Quality Report?
▸ The quality report (.pdf format): …\project_name\1_initial\report\project_name_report.pdf

Could you also tell us more about how did you process your dataset? You mentioned performing several flights. Did you divide the imagery per subprojects; oblique, nadir and merge them at the later step?

The more we know about your workflow and data the better we can assist you :wink:

I look forward to your response.

Best!

Hello Beata,

Thanks for your quick response! When processing my dataset, I manually isolated the higher elevation photos and imported them (490 images). I then input GCP data, marked images, and processed the point cloud and mesh. I did not separate nadir or oblique images, as they were all in the same folder. If this is a step that needs to be taken, I am unsure how best to do it. My previous experience with photogrammetry has not required such divisions.

I have attached my quality report as well. Thanks again for the help. I hope that I have just made a simple mistake!

Cheers,

DillonCrowninshield HE Report.pdf (103.9 KB)

Hi @scanning,

You quality report shows strange things.

First, it says that there is no geolocation and no GCP.
Have you marked your GCPs?

**

**

Are you sure that your images have been acquired in this coordinate system?
NAD83(2011) / Delaware (ftUS)
Have you change it? Is so, I wouldn’t.
Why did you set the output coordinate system as Arbitrary?
Here I would select this one:
NAD83(2011) / Delaware (ftUS)

Regarding the processing options, I would select fro the Calibration method:
“Accurate Geolocation and Orientation”

Could you give a try and let us know if it improves a bit or not.

If not, could you please share with us your data set?

Thank you,
Best

Marco,

I ran this project a few times, attempting to change one variable at a time to understand what was causing my issues. I did indeed mark GCPs and assigned their appropriate coordinates based on DE state plane. I am unsure why the report shows none…

I am unsure what coordinate system the images were captured in, but P4D automatically detected NAD_1983_StatePlan_Maryland_FIPS_1900_Feet. In one variation I selected the NAD 2011 DE State plane, in another I selected the arbitrary. My hope here was that the images would align as a model that could later be assigned geo located points.

To further expound on this challenge, when flying the site with our P4RTK we were unable to attain any type of RTK lock with our D-RTK 2 pole. The image data is therefore less reliable.

This really is outside of my normal workflow, and I plan to take some courses, online or otherwise to better understand how this software works.

Here is a link to my data set.

Hi Dillon,

First of all, Happy New Year!
I was intending to download the dataset, but the data transfer session expired. Could you please upload the images again on our OneDrive?

Best wishes,
Teodora

Teodora,

I have uploaded all images acquired. I hope you are able to make sense of what is going on!

Thanks for your help.

Dillon

Hello again,

What are the exact 490 images that you used in this project? Please mention the image filenames, because I do not know which of the 2 uploaded folders to choose.
Also:

  • What is the output coordinate system you wish to obtain in this project?
  • Can you also provide the file with the GCPs (with info about their coordinate system)? I would like to replicate this issue and investigate thoroughly.

Best,

Teodora

@scanning -
“I ran this project a few times…” is probably why those GCPs that you marked aren’t showing up. In order for the model to incorporate your GCPs, follow these steps:

  1. Run step 1, initial processing.
  2. Mark your ground control points in the raycloud editor. (If images are not showing up at this point, when you click one if your GCPs in the menu on the left side of pix4d, you can mark them using the basic GCP editor from the map window or the file menu. This is usually because people are using DJI drones and the image altitude is significantly different than the GCP altitude. )
  3. (This is the important part) Select “Reoptimize” from your processing menu to incorporate your ground control.
  4. Now you have GCPs, unless you re-run step 1 again…at which point you just un-did all that work marking ground control points. I’m guessing this is what happened.

Regarding coordinate systems:
Never change the coordinate system for your IMAGES. You have 3 places to change your coordinate system:

  1. Image coordinate system (leave it alone)
  2. GCP coordinate system (ask your surveyor. Ideally, get a EPSG/WKID/FIPS whatever code.)
  3. Output coordinate system (ask your surveyor and/or your client)