Issue with thermal band alignment

Hi Community, I am experiencing a thermal band alignment issue in ODM.

I am using PyODM to process MicaSense Altum multispectral images, including the thermal band.

Due to the thermal camera’s position, the input thermal bands are misaligned to the left with respect to the other 5 bands, as you can verify with a normal viewer by comparing the various bands (on a single shot) (Google Drive’s link: I have inserted both the original bands and those resized using an appropriate script).

The problem is that, although in the log file (see ‘task_output.txt’) reference is made to an alignment operation of the LWIR band, this misalignment is also preserved in the orthophoto generated by ODM, contrary to what happens for the remaining 5 bands (to speed up the test, in addition to the thermal band, I gave to ODM only 1 different band instead of 5, but the issue also occurs with 5 bands ): see screenshot ‘ODM_thermal_band_offset.png’, in which the lighter rows represent the thermal band of the ODM orthomosaic and the darker ones in green the reference rows of the Metashape orthomosaic (viewable below in transparency).

Among the many tests carried out, I point out three important ones, in which I used the following input options:

  1. dsm: true, dtm: true, min-num-features: 64000, crop: 0, texturing-skip-global-seam-leveling: true;
  2. dsm: true, dtm: true, min-num-features: 64000, radiometric-calibration: camera + sun, crop: 0, texturing-skip-global-seam-leveling: true;
  3. dsm: true, dtm: true, min-num-features: 64000, crop: 0;

In all cases, the spatial shift was not eliminated, compromising the ability to correctly read the temperature values ​​in correspondence with the rows of a vineyard.

I performed both processing with the original input images (160x120 resolution) and with the previously resized images and brought to the resolution of the other bands (2064X1544): in both cases, I found the spatial offset of the thermal band in the final orthomosaic.

In Metashape, on the other hand, the same input dataset, processed using the default parameters, produced an orthomosaic that does not present this problem on the thermal band (see ‘metashape_no_thermal_band_offset.png’).

Is this a known problem in ODM or do I need to pass to it some specific options that I don’t know?

Thanks in advance for any support!

Google Drive’s Link: ‪Thermal_Alignment_Issue‬‏ - Google Drive

1 Like

Welcome!

Sorry for the trouble!

It looks like your process/parameters are pretty much spot-on, so this might be a limitation or implementation issue in our pipeline.

Are you able to share your full test dataset?

Hi Saijin_Naib,

Thank you for your quick reply!

At this link you can find my full test dataset (0017SET.zip):

https://drive.google.com/drive/folders/1UazdTKIxyiQfvNg1oK_rsXSFBuuAeqle?usp=sharing

Looking forward for your feedback, thank you in advance!

1 Like

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.