Pix4DMapper Orthomosaic Shifting - Pixel/Resolution Shift

PIX4Dmapper PIX4Dmapper Questions/Troubleshooting
1

I’ve noticed over my time of using the software that there is a shift showing up between my orthomosaic and ground control points that does not exist in my point cloud. This doesn’t appear to be affected by units, coordinate system, arbitrary, etc. and my RMS error of control always comes back satisfactory (1-2 cm). Typically, i used the ortho solely for a visual reference and wasn’t concerned of the minor error as the map scale i used made the error negligible. Now i am looking to use the imagery more for location digitizing, and sharing with others, so i would like to figure out how to remove this shift.

After paying closer attention, the shift appears to be directly tied to the pixel resolution i choose for the map. I typically lower the resolution from standard GSD (1-2 cm) to (7.62 cm/0.25 ft) for better data size management. I notice, the imagery is being shifted to the Northwest by approximately this resolution distance (+/- the survey error). This shift is consistent with all control points in the same direction.

To me, it appears the 3rd party software I use, could potentially be reading the raster from a different location on the pixel? This error shows up in both Autodesk Civil3D as well as Trimble Business Center software packages. I searched all over to find a setting in both 3rd party software, as well as PIX4D Mapper to find a fix for this issue and have had no luck.

Is there a way to remedy this within the PIX4D Mapper Software?

Thanks in Advance,
Joe

OrthoShift
OrthoShift570×630 32.2 KB

2

Hi Joe,
If you can provide the PDF quality report, we can take a look to see if there are any processing issues.

However, considering the shift is constant between all of your control points, it tends to suggest a discrepancy in the units. Specifically, if you process using US feet but set the map units in Civil3D to International feet, you would see this type of shift. Can you take a close look at this?

3

Hey Mike,

I’m in the field today but will upload a processing report for you to take a look at tomorrow.

It is definitely not a units conversion issue. I’ve had problems in the past that some software misinterprets pix4d’s label of “feet” and not international/us survey feet, and now I process all of my data in meters. Plus the state plane system and zone I work in, the difference is approximately 50’ on Easting which is easy to catch.

As mentioned, it almost appears how the Geotif is written and associated with location of coordinate to the pixel.

Thanks,
Joe M

4

Hey Mike,

Processing report attached. Let me know if anything stands out to you that i may be doing wrong.

Thanks,
Joe

2024-07-18_PrincetonPit_report.pdf (3.2 MB)

5

Mike,

Following up again to see if there is any resolution to this issue?

6

I’m having the same problem. The exact same problem.

Is there any resolution?

7

I have never received a response/solution to this issue. It’s actually been pretty difficult to get useful (or quick) feedback/resolution on this forum.

I suspect a couple things could be the cause;

  • I am using a phantom 4 RTK. I do note that the default camera calibration settings provide a principle point that isn’t the actual image center. I expect this would get resolved during calibration, but i am not sure that if it would cause a similar issue. I don’t quite have the depth of knowledge to feel comfortable moving away from the recommended camera settings (shy of a focal length calibration).
  • I also believe there could be a difference between how PIX4D defines a GEOTIF grid, and how the software that I use (primarily Trimble Business Center and AutoDesk Civil3D) interprets it. I get curious if one defines based on the pixel center, while the other may read based on one of the pixel corners. The shift always appears equal in distance and direction between points. Once the images are brought into Civil3D, I can actually shift the imagery equally and it aligns rather well (with some error more in line with the control adjustment). This works fine for my internal workflows, but will not suffice for the product that I provide to my clients if they require Ortho imagery.

These are just ideas and are likely not accurate, but its what I’m left with as I cannot get a response from PIX4D…

I am actually in consideration of switching platforms to Agisoft Metashape due to the poor customer support and lack of software updates to better support datums in the US (and of course international feet). I plan to perform a trial in the upcoming months and will share my findings.

If you come across a fix for this, please share!

8

When importing a .tiff file into Civil3D, the GeoRefImg command must be installed. Otherwise, you will experience a shift in the outputs. You can read more in the article below.