Leica ADS100 support?

[StevenF]

Would Metashape be able to create dense point clouds from Leica ADS100 images?

This an aerial digital pushbroom system with forward, nadir, and back looking sensors used for stereo capture. The data comes in SOCET support file format with a number of separate TIFF files (called blocks) forming one image. There are a number of corresponding auxiliary files including a camera calibration file (.cam) and exterior orientation stored in a (.odf) file.

Metashape supports satellite pushbroom sensors with RPC (rational polynomial coefficient) info in a text file. If I merged the TIFF blocks and then created an RPC text file from the existing auxiliary files, would the data work correctly in Metashape?

[StevenF]

To anyone else who is curious the answer is yes. If you merge the TIFF files of ADS100 images and produce RPCs in other software then Metashape will use them correctly and produce good results for processes like dense cloud generation.

The challenge I'm facing now is that the images are so big that parameter testing with even a single pair takes a long time. Resizing the region and masking images doesn't appear to save time in depth map generation.

[Alexey Pasumansky]

Hello StevenF,

How big are the source images? Would it be possible to share them and project with the alignment results, so that we could check the depth maps generation using smaller region or with masks?

[StevenF]

Hi Alexey,

The images are about 310k pixels by 20k pixels and 16 bit. They're about 17-20 GB per image. I tried depth map generation with a smaller region and mask but it still takes about a day for a single pair.

Let me know how I can send you the project and example images.

P.S.
Sorry I didn't see your reply here and started a new dedicated thread, which you can ignore if you want.

[StevenF]

I discovered that I can clip a TIFF with RPCs using gdal_translate -srcwin. GDAL outputs a TIFF file with the corrected RPC tags. The only change is in the line offset and sample offset.  The clipped images can then be used to test parameters in Metashape over a smaller area than the full images.

If there's a way to generate depth maps over only a portion of an image, that would be more convenient, but this work around is pretty simple.