M3E Camera Calibration

[DOM112]

Hello Fellow Mavic Pilots,

We use DJI products extensively for mapping and surveying. We recently upgraded from the Phantom 4 Pro V2 to M3E with an RTK module. We purchased 2 M3Es in the past few months. One seems to be working well while the other seems to have an improperly calibrated camera. We use Pix4D for processing. The initial camera settings are correct for the M3E. The Pix4D quality report states the the relative difference between initial and optimized internal camera parameters is 20.64%. We normally see less than 1% and often less than that if use PPK GEOTAgs. Camera Optimization is tagged red in the quality report, which is the only indication something is askew. Everything else looks normal throughout the three processing steps. We used four GCPs with a .036 US survey foot error. However the elevation data on the upper contours is off by approximately 18ft. Never thought this was possible with GCPs, but we have irrefutable evidence this is indeed true.

We reprocessed the data set with the optimized focal length parameter from the Pix4D report. While the camera optimization went from 20% to 1% and was tagged green, the elevation error persists. Retrospectively, this isn't surprising. We also processed the images post kinematically with high quality and nearby base data. This didn't help.

Has anyone seen anything like this? Is there a way to calibrate the camera on the M3E? We could do this on the P4Ps but I don't see method to do it in DJI assistant 2

Any insight or suggestions are appreciated.

I attached a redacted quality report.

Thanks,

DOM112

 

[yarrr]

That is interesting. With the optimized focal length being longer- is there a possibility you had "zoomed" in the camera for the mission? Also strange that the 'dewarp' deadspots aren't showing up on the top of ATP per pixel graphic.

You can save the optimized parameters in Pix4d, which is somewhat of a calibration for the purposed of post processing, but I'd be curious to know if an actual/true lens calibration is available from DJI or third parties.

Seems to me that the vehicle is not functioning correctly for its intended purpose, and should be exchanged though.

 

[DOM112]

That is interesting. With the optimized focal length being longer- is there a possibility you had "zoomed" in the camera for the mission? Also strange that the 'dewarp' deadspots aren't showing up on the top of ATP per pixel graphic.

You can save the optimized parameters in Pix4d, which is somewhat of a calibration for the purposed of post processing, but I'd be curious to know if an actual/true lens calibration is available from DJI or third parties.

Seems to me that the vehicle is not functioning correctly for its intended purpose, and should be exchanged though.

We considered the possibility that the mission was flown with a zoomed in lens. The EXIF data on the images indicate the focal length was 12.2792. Empirically, the operator claims he didn't zoom in.

We did a bit more testing yesterday. We took images from two M3E at the same altitude (200ft) and relative position. In the attached images, you can clearly see differences in the images. We are fairly certain the lens on the new M3E is improperly calibrated or defective. Without vanishing points, I don't think I can calculate the focal length. Topodrone, which is what we use for PPK, has a calibration tool, but I am not sure how to use it.

What do you mean be dewarp deadspots? Is this something I should be looking at in the quality report?
Thanks for your reply,
DOM

 

[Arnold LeVine]

The barreling (4220) is terrible. the other photo (4440) is very rectilinear.
Send it back to DJI.
The color shift between shots is dramatic. I don't know which is more accurate.

 

[DOM112]

Well we did have a set up issue that contributed to some of the problem. We didn't have dewarp off nor did we have mechanical shutter selected. We did another flight with the recommended mapping set up and compared it to a flight with a known good M3E. Same altitude, same camera options and same processing conditions. Again we are seeing a significant divergence between initial and optimized focal length for the bad drone. The good M3E has an optimized value with .15% of the initial value, which sharply contrasts with 7% for the bad drone. I've attached both quality reports and static images taken from 100ft with both drones.

Seems pretty clear the camera is improperly calibrated for defective.

Thanks,

DOM

 

[DOM112]

Well, we heard back from DJI Customer Support. To my dismay and disappointment, they claim the drone is fine. They told me to contact PIX4d. Not sure how they reached their conclusion, but I have compelling evidence something is wrong with a brand new M3E. Never crashed or handled roughly. This is ********!