A lot of uncalibrated cameras


I’m writing regarding an issue that I have with processing an area of forest from a drone flight.
I get a lot of uncalibrated cameras, therefore unable to have a proper orthomosaic output (there are is a high area missing in the final output).
I tried several options found on pix4d community:

  • reducing the image scale
  • using Alternative Calibration Method
  • set the Internal Parameters Optimization to “All Prior”
  • reducing the number of Keypoints (first to 40000, then 35000, then 25000)
  • using the Triangulation method instead of the Inverse Distance Weighting (IDW) to generate the DSM

Unfortunately, none of the above worked.

The flight was performed at 100 m height, with a Horizontal Overlapping Rate of 70% and Vertical Overlapping Rate of 80%.
As an additional information, note that the flight was performed during early spring and therefore the trees were not green at all, the forest was actually dry.

Can someone please advise on how I can process this flight and generate a proper orthomosaic. Being just a forest, I’m not looking into having a high quality or precise output. I just need it to be generated in a way just to see that there is a forest, even though the image quality wouldn’t be high.

Attached, is the quality report.

Many thanks in advance!

Urleta 24-1_report.pdf (2.0 MB)

Hi Cristi,
It appears that the elevation of the project area rises in the center. Is that correct? If so, then your overlap will be lower as the elevation rises. Insufficient overlap might be the cause and not the vegetation itself. The best solution would be to refly the area at a higher altitude or use a terrain following flight application.

Hi Mike,

Thank you for your reply!
The elevation area does not rise in the center, actually. Could the height of the trees give this impression?


Hi again Mike,

After trying several processing options and using different parameters I found the following options that worked for me and resulted in 94% of cameras calibrated, which significantly improved the output:

Image Scale: 1/2
Calibration method: Accurate Geolocation and Orientation
Internal Parameters Optimization: All Prior
External Parameters Optimization: All
Rematch: Custom-Rematch

Note that I used these options on other projects with dense vegetation as well and I got the same good result.
Maybe this will also help other people encountering this issue. I saw many discussions on this topic.

For reference, I attached the new Quality Report
Have a good day!
Urleta 24-1_report.pdf (2.1 MB)