Camera calibration flight?

PIX4Dmapper
1

Hi all,
So we have a DJI M300 RTK with a P1 camera but we have recently been having some issues.
Due to not having calibrated camera parameters specific to our P1 we have been leaving “Internal parameters optimisation” as “all” in initial processing.

This has worked perfectly for some flights but for others the point cloud has come out with errors of between 0.3m to 0.6m in elevation. Image precisions are RTK (<50mm XYZ)

The current work around for this has been to copy the optimised camera parameters from a flight that produced very good data (<0.1m vertical) and setting “internal parameters optimisation” to “none”.

This has worked well so far but I am wondering if there is a process to conduct a flight with multiple GCPs or something to better determine these parameters without sending the camera away for lab calibration.

Thanks

2

Hi rsaunder,
If you are seeing high camera optimization errors (ie. a yellow check mark in the quality check) then you should enable All Prior. This will force the optimal internal parameters to be close to the initial values. Basically, Mapper will trust the camera model more than it otherwise would. Give this a shot and if you still have problems let us know.

3

Thanks, although the issue seems to almost be the opposite.
Pix4d optimises the camera very closely to the original parameters (0.04% or so usually) however these are the flights where the point cloud is showing height error.
In other flights the optimisation is more like 0.2% and these have proven to yield much more accurate data.
Trying to find out if there is a way to calculate a better set of default parameter to save into the camera DB

4

If youre finding that the optimized parameters are working better than the database parameters then you should consider saving the model and you can just load that instead of copy-pasting. As long as you are confident in your model then you can set the internal optimization to all prior and just have that as the new base. If you are experiencing error still then its likely related to something other than the model.