Situation: I collected around 2000 drone images over 200 ha. My computer does not have the resources to render one large orthomosaic at once, so I decided to split my dataset into 8 and produce 8 different orthomosaics. I am using WebODM for my image processing. Because I did not use GCP's, all my 8 orthomosaics do not align properly with one another, and I want to merge them into one large orthomosaic.

Question: What software/technique can I use to align my 8 orthomosaics so I can merge them into one?

I have tried: I found this Python library arosics (https://pypi.org/project/arosics/) to co-register overlapping imagery. However, I cant get it to work with my drone orthomosaics. I believe it has to do with the pixel type, but not sure. I can share the code and output here if someone is interested.

I know cloud processing is a solution, but I do not have the funds at the moment to make this a viable solution.

  • I have resolved the arosics error. Had to do with No_Data values. Other alternatives to this problem is still welkome – kjtheron Oct 3 '20 at 7:29

Structure from Motion (SfM) software packages like WebODM are using the available images to piece together an orthomosaic. On the outer edge of the mosaic there are fewer images for the algorithms to choose from. You may find, now that you have 8 different orthomosaic images, that you have an impossible 'apple and oranges' situation where the overlapping pixels are stretched, selected, and distorted in different ways.

One solution might be to use something like QGIS (or a GIS related Python package like Shapely, Pyshp, or GDAL/OGR) to make polygons smaller than the extent of each image but with no gaps between the polygons. Clip the images to their respected polygons. Merge the resulting rasters together or place them is a single directory and create a VRT from the images.

  • Thanks for your suggestion. However, it will not work. GPS of the drone I used is not very precise (like most hobbyist drones), and the GPS error is averaged per orthomosiac. Between orthomosiacs the GPS error is averaged differently, and hence the orthomosiacs will not linr up accurately. Also, WebODM already crops/clips 3m of the edge of the mapping area – kjtheron Oct 7 '20 at 14:32
  • Agreed about the GNSS error of a hobby-grade drone. Still, you could use QGIS to georefference the resulting orthos to some existing imagery or pick one of your orthos and georeference the remaining images to that image. And, while I have never tired it - what happens if you feed your 8 images into a new WebODM project. Will WebODM stitch those images together? – GBG Oct 7 '20 at 15:32
  • I found the georeferencing function within ArcGIS Pro, and will use this to align my orthomosiacs. However, WebODM have added support to TIF imagery, so if I can embed the image headers with data, a SfM workflow in WebODM could work? Will play around and report back here. – kjtheron Oct 7 '20 at 20:26
  • The orthos coming out of WebODM are geoTiffs so you don't have to embed the images headers with spatial data. – GBG Oct 7 '20 at 22:14
  • WebODM is unable to load the orthophotos. Guess the best is to georeference them. Thanks for your input @GBG – kjtheron Oct 8 '20 at 14:19

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.