What causes Camera Location error?

[iaerobotics]

I had a Rock Robotics R3Pro Lidar system. Its an Epic EasyScan 10 re-brand. It used to generated the most accurate Ortho's I have ever seen. It would exactly match the LiDAR data and need no shifting or rectifying. This was the only unit I had ever had that was ever that close.

Fast forward to a crash with the unit. I bought a new unit, same model. This model does not produce nearly as accurate Ortho's when processed. It is especially noticeable in non-overlap straightline projects.

I have attached 2 "pre" reports which is from the old unit and 2 "post" reports from the new unit. You can see the old unit had errors on x,y in the mm range and the new unit is in meters.

What can cause this? I am not sure what drives these errors. Is it camera location in exif compared to calculated camera location after processing?

Do I look to see if there is some lens issue or calibration I need to do within Metashape that is causing the error? Is it accuracy on the images when they come into Metashape? What is this report telling me when 1 is accurate to mm and the other at meters?

Thanks in advance.

[Paulo]

Hi aerobotics.

for one line survey, it is critical that you have precise orientation angles for your cameras. So I would suggest that you check the angles box before aligning your cameras. Do you have txt file with the cameras' eo parameters (location and rotation) from your GNSS/IMU processing? Is it accurate?

[iaerobotics]

Yes, there is very little to no difference when I import the exact IMU angles. The previous unit with just the exit info on gps location was more accurate than this unit with all the exact camera angles included. I just don’t understand what is reported in that report item for accuracy. It just stood out that the previous unit was in the mm range and this one is in the meter range.

[Paulo]

Ok,

please redo an alignment with both location and angular reference box checked and you should get a good result...

Or if you can share small dataset then I can have a look at it...

[olihar]

What does Rock Robotics support say?

[iaerobotics]

What does Rock Robotics support say?

Haha. What they always say. “If it does work, fly it again. If it still doesn’t work we are not sure why. Sorry”

They just rebrand it and support is at a very basic level.

[iaerobotics]

Ok,

please redo an alignment with both location and angular reference box checked and you should get a good result...

Or if you can share small dataset then I can have a look at it...

https://drive.google.com/drive/folders/17c5Edl5gINSCDiYOsbHjgHutSmKeV2hD?usp=sharing

[iaerobotics]

Hi aerobotics.

for one line survey, it is critical that you have precise orientation angles for your cameras. So I would suggest that you check the angles box before aligning your cameras. Do you have txt file with the cameras' eo parameters (location and rotation) from your GNSS/IMU processing? Is it accurate?

Enabling this did move it into the cm error which is much better than meter. Still not in the mm range it was before and I still don't understand why I have to enable this YAW checkbox when I didn't before.

[Paulo]

Hi again,

i was able to process the images using the camera positions and orientations calculated by PPK. The result is very good with errors at 1.5 cm in XY and 5 mm in Z. This is normal as the ppk processing has errors as well as off set s from gps to ins to camera...Find report in https://drive.google.com/file/d/1780_2pHl1k9KK8bINCGTfZnkJIVVlnpX/view?usp=sharing.

If you need more info, please contact me by PM.

[iaerobotics]

Hi again,

i was able to process the images using the camera positions and orientations calculated by PPK. The result is very good with errors at 1.5 cm in XY and 5 mm in Z. This is normal as the ppk processing has errors as well as off set s from gps to ins to camera...Find report in https://drive.google.com/file/d/1780_2pHl1k9KK8bINCGTfZnkJIVVlnpX/view?usp=sharing.

If you need more info, please contact me by PM.

It was in the mm range before. I can send you another one that even with this checked it's almost 2m in error.

[iaerobotics]

Hi again,

i was able to process the images using the camera positions and orientations calculated by PPK. The result is very good with errors at 1.5 cm in XY and 5 mm in Z. This is normal as the ppk processing has errors as well as off set s from gps to ins to camera...Find report in https://drive.google.com/file/d/1780_2pHl1k9KK8bINCGTfZnkJIVVlnpX/view?usp=sharing.

If you need more info, please contact me by PM.

New data up there- same link. Would love your thoughts on that one.

[iaerobotics]

Interesting point - that lends me to believe the issue is less accurate geotagging of images with the new unit is that if i reduce the accuracy from highest to HIGH, the error gets less and when reduced to medium the error went from 5m at highest to .5m at medium. Hmmm.

[Paulo]

New data up there- same link. Would love your thoughts on that one.

Hi John.

works fine with me...After initial alignment 9without optimizing. I get very small camera location errors. See attachment...

By the way, what is the model of your GPS base receiver?

[iaerobotics]

New data up there- same link. Would love your thoughts on that one.

Hi John.

works fine with me...After initial alignment 9without optimizing. I get very small camera location errors. See attachment...

By the way, what is the model of your GPS base receiver?

Odd. Can you export the full report so I can see what processing options you used? After initial alignment I get 5m error without rotation checked and 1.9m with it checked.

[Paulo]

The final optimized chunk with imported laser scan and generated point cloud is shown in attached image screen.   What did you define as your camera position accuracy?