Large errors for yaw, roll and pitch?

[Lambertus]

Hi guys,

Currently I am working with aerial images trying to create a point cloud from a certain area in my country.
I imported my camera positions including yaw, roll and pitch. However, it seems that including yaw, roll and pitch increases the error in my model. See the attached image for more information. Should I exclude roll, yaw and pitch from processing maybe?  Thanks in advance for your help!

[Alexey Pasumansky]

Hello Lambertus,

Orientation angles are not included in the referencing procedure according to your screenshot (see the check box in front of Yaw angle in each row).

As for the reason of the 180 degree errors - it may be a result of image rotation in the external editing application (if the image is opened, rotated and saved) prior to the loading the images to Metashape. Or the angles are not properly recorded and saved in meta information.
Do the loaded Yaw angles properly correspond to the flight direction? 0 degrees in one direction and 180 degrees in the opposite?

[Lambertus]

-

[Lambertus]

Hi Alexey,

Hello Lambertus,

Orientation angles are not included in the referencing procedure according to your screenshot (see the check box in front of Yaw angle in each row).

Thanks for your reply. I didn't know that I needed to check the boxes in front of the yaw column. I will try if checking these boxes will increase the accuracy of my model. 

When I loaded my raw aerial imagery into PS, I saw that some images were rotated and some not.  Outside PS I rotated all my images 90 degrees to the right, clockwise direction. After this I added my photos to PS. It seems that all images were positioned right yet. 

The camera position data including yaw, roll and pitch is not related to my manual rotation action. These yaw, roll and pitch values were attached in separate files to my aerial images. I combined them into one file and added this via Reference Pane.

Do the loaded Yaw angles properly correspond to the flight direction? 0 degrees in one direction and 180 degrees in the opposite?

As far as I know, these yaw angles correspond with the flight direction.

I have another question as well. I saw the reprojection error in the info of my chunk. Does a large error mean that a produced point cloud will totally mismatch when draping this point cloud on top of a correct aerial basemap?
(assumption: point cloud and aerial basemap have the same projected coordinate system)

[Alexey Pasumansky]

Hello Lambertus,

Can you share the project re-saved after the alignment in PSZ format with the thumbnails available with our support (support@agisoft.com)? I'll take a look on the alignment results to see, if there might be anything wrong with the yaw angle? In case you can also share the auxiliary files (at least a subset) related to the exterior orientation data it may be also helpful.

[Lambertus]

Thanks for your quick reply.
I will share my project in PSZ format including thumbnails. Furthermore I will send my csv file with orientation data.

[Alexey Pasumansky]

It seems that most of the images are 180 degrees rotated (for some reason) which results in 180 degree errors compared to yaw estimation. For  this dataset the issue could be fixed by using 180 degree offset for yaw angle in the Camera Calibration dialog. But there are a few images with the proper orientation, then their angles should be unchecked in the Reference pane after this correction applied, otherwise their "corrected" yaw would affect the results.

[Lambertus]

Thanks for your reply. It is clear to my now
I managed to fix the issue using the GPS/INS window in the Reference pane.