Alignment error when processing chunk of images with GCPs along one axis only.

[March Castle]

I am having issues with certain datasets when a portion of the images won't align.

I split the un-aligned images into a separate chunk and align them with the aim of aligning and merging the chunks together.


Unfortunately, after alignment the images appear tilted to the side in relation to the other chunk, and after aligning and merging the chunks, the tilted data remains (see attached)

Can this be solved, or is this caused by having a single line of GCPs along the edge of the dataset which has a strong bias along one axis?


I have tried reprocessing this, but the same issue persists whether I identify the GCPs before or after processing.

Thanks for your help.

[Alexey Pasumansky]

Hello March,

If you have the camera orientation angles available, you can check on them in the Reference pane to correct the model orientation in the coordinate system space.

[March Castle]

Unfortunately we are using images from Sony cameras which don't have any GPS information to assist with alignment.

We rely purely on the software to align the images and then location data is provided by identifying the GCPs.

The position of the GCPs are logged by a Trimble Geo 7x GNSS unit which provides centimetre accurate positioning.


This error happens in several of our datasets - most of the images align well, but a section of the images will be aligned at an extreme angle.

These sections contain aligned images, but are all projected at the same incorrect angle.

Reprocessing produces different results, but they are always nearly upside down.

I have attempted to process these sections by moving the affected images to a new chunk and resetting the alignment, the aligning and merging the chunks, but this produces the same results.


I think the content of the images presents an issue for Metashape - in that we fly over plots of vegetation which are numerous and quite similar. Although the pattern we fly is waypoint-based and so each image will always be spatially linked to the images taken before and after it., which I assume would make it more straightforward?

I have tried different settings in the alignment stage: 'Source' and 'Sequential' but this seems to make little difference.

Even when the two misaligned chunks share GCPs which have been identified, the alignment still fails.

Thank you for your help with this issue - I would appreciate any suggestions!

[Paulo]

Hi March,

in case the misaligned tilted part only has GCPs along a line. then tilted nature can be explained by this pattern. My suggestion is to create at least 3 GCPs (not on same line) in correct chunk that overlap with tilted chunk. And identify and measure these same markers in tilted chunk. Then align chunks using marker based method...

Should work...