RedEdge-P orthomosaic band misregistration, but same dataset works in Pix4D

[ahornero]

Hi,

I am processing a MicaSense RedEdge-P dataset in Metashape Professional 2.3.1, and I cannot obtain a correct multispectral orthomosaic.

I have processed RedEdge-P datasets in Metashape many times before using the usual multispectral workflow, normally with the panchromatic band as the master band, and I usually obtain good results. However, in this particular case, the orthomosaic is clearly not usable.

The project seems to align correctly: 954 images, 936 aligned, and a reprojection error of around 0.334 pix. The orthomosaic is generated as a 6-band uint16 product, but the bands appear mixed or spatially misregistered, producing strong colour artefacts across the whole mosaic.

I have already tried:

• using the panchromatic band as master;
• generating the orthomosaic without the panchromatic band;
• reprocessing the project from scratch.

In all cases, the result remains unacceptable.

What is confusing is that the same dataset processed in Pix4D produces a correct orthomosaic, so the issue does not seem to be related to the raw imagery itself. This makes me wonder whether there is a specific Metashape setting, a RedEdge-P / panchro workflow issue, or possibly some version-specific behaviour in the latest release.

I have attached the Metashape output previews from two attempts, the Pix4D output for comparison, and the Metashape processing report, compressed due to the forum size limit.

Any suggestions on what to check next would be very helpful.

Many thanks.

[Alexey Pasumansky]

Hello ahornero,

Can you please send the project with the alignment results only saved in PSZ format to support@agisoft.com? So leave only aligned camera locations, tie points and markers, if any, in the project copy and save in PSZ format.
Also please share the PDF report exported from the original fully processed project.

In case the bands are mixed up (as you see them in Camera Calibration, Calibrate Reflectance and Raster Calculator dialogs) it is likely result of incorrectly added source data. Have you added all the images at once using Add Folder -> Multicamera layout option?

[Paulo]

I have also processed a 6 band (Blue,Green,Panchro,red,redEdge,NIR) Rededge P data set.

Alignment works fine see following Tie Points Info:

Code: [Select]

Points 165,189 of 698,669
Point colors 1 bands, uint16
Key points No
RMS reprojection error 0.186576 (0.303193 pix)
Max reprojection error 0.517046 (3.13311 pix)
Mean key point size 1.56237 pix
Average tie point multiplicity 9.69289
Alignment parameters
Accuracy High
Generic preselection Yes
Reference preselection Source
Key point limit 40,000
Key point limit per Mpx 1,500
Tie point limit 5,000
Filter points by mask Yes
Mask tie points No
Exclude stationary tie points No
Guided image matching No
Adaptive camera model fitting No
Matching time 1 hours 46 minutes
Matching memory usage 2.10 GB
Alignment time 19 minutes 18 seconds
Alignment memory usage 1.12 GB
Optimization parameters
Parameters f, cx, cy, k1-k3, p1, p2
Adaptive camera model fitting No
Exclude corners No
Optimization time 23 seconds
Date created 2026:05:21 16:47:54
Software version 2.3.1.22446
File size 108.61 MBThen I generated a dense cloud, DEM and ortho. But when I use different formulae to display output, I get weird results.

For example, if I try to display in false color (NIR/32768, RED/32768, Green/32768), the dense cloud displays in false color but not the ortho (greyish tones) and obvious image misalignments in mosaic. See 1st attachement.

If I use Panchro band/ 32768 to display, image misalignment is even worse, see 2nd attachment.

Obviously we would need some guidance to coreectly generate and sisplay ortos with this type og imagery....

[virtek]

I believe there is another problem: during data import, when the 'multi-camera system' option is selected, the bands do not load properly (the first band is loaded and then shifted as if in a matrix). Consequently, the raster calculator cannot be used on the orthomosaic. this happens on two computers with version 2.3.0-2.3.1 (drone DJI Mavic 3 Multispectral)

[Alexey Pasumansky]

Hello virtek,

Do you load only TIFF files coming from multispectral sensors when using Multi-camera System approach or are trying to add both TIFFs and JPEGs?

[rival29]

I am having the same issue with the RedEdge-M Dual. I had to go back to 2.2.3 to get these to process correctly, I tried everything! Camera Calibration, precalibration, gnss offsets and bias, slave adjust on off, group separation, chunking.... 2.2.3 is working solid, the only feature im missing a bit from 2.3.x at the moment is the network level tasking so I dont have any workers waiting around. Confirmed on our rededge 10band.

attached image, i turned off mosaic blending just to confirm that wasn't part of the issue. Its not handling the band alignment correctly

[Alexey Pasumansky]

Hello rival29,

We have just updated 2.3.1 release version to build 22580, can you please check, if the reported problem is resolved now?

[rival29]

absolutely, is the newest version retrievable in the downloads area on website?

[rival29]

Confirmed fixed with new version!!!

[Alexey Pasumansky]

Hello rival29.

Glad to hear that the issue is resolved. We apologize for any inconvenience caused by the bug in the previous release version.