A while back we took a look at when and why alignment in Metashape fails. In this blog we will take another look at how alignment issues can be triggered and overcome.
In common with the previous example we were asked to review a set of images covering a transect of the Great Barrier Reef in Australia.
This time the primary issue was a failure to align. Overlap looked good but there was a gap between chunks that refused to cooperate and a deeper look revealed there was a patch of reef images that really lacked key points.
Lack of key points really messes with alignment in Metashape. We either lack enough images for robust alignment, poorly distribute them, or face both issues. Either way, during recursive optimisation (something we teach in our Metashape Professional training course) the removal of inaccurate tie points can see images removed.
After all, if the image is not aligned and position accurate its inclusion is unwelcome.
The issue arose from fluctuating light levels caused by wave action refracting sunlight as it entered the ocean. The ripples of light on the sand were affecting how Metashape examined the images, both during preselection phase and key point detection.
Shooting the images on a cloudy day would be perfect. But, perfect does not always exist in the real world. So, how do we work around this issue?
Generic Off – Guided Matching On
The first step to fixing alignment in metashape is to turn off Generic Preselection. Normally this feature is really desirable as it helps with key point matching by creating a set of visually matched image pairs beforehand.
But with fluctuating light levels then image matching might not be so good…solution is to turn it off.
Next we can turn on on Guided image matching. This feature takes a deeper look into the images and boosts the number of key points found. Consequently, we recommend using it for vegetation and other high-resolution source images.
The results speak for themselves:
The difference in the number of feature and tie points was profound…and a sound result. The only downside is alignment processing times increased in Metashape. However, it was a lot quicker and cheaper than sailing out to the reef and re-running the transect.
With robust alignment, we could recursively optimize the image bundle to refine the aligned images before adding scaling.
The Results
The result is a visually pleasing, scaled and accurate model with an ortho photo processed at 1mm per pixel:
As ever, the ortho photo is hosted by our friends at Dronelab. Be sure to zoom in and check out the detail in the corals.
We cover much of alignment, what we discussed in this blog, in our online metashape training. Right now we are offering our training courses with the option to see 50& of the course fee. Donated by AccuPixel to a charity delivering humanitarian relief in Ukraine.
If you are thinking about taking one of our courses…do please consider the option of providing some help to others at what can only be a desperate time.
For the more challenging use cases…the good news is both Jose and Simon relish in solving problems. Consultancy work can involve everything from refining or validating a workflow, improving accuracy and repeatability or creating custom python scripts to automate or extract more value from existing data.
Please get in touch if you think we may be able to assist your project.
Closing words on alignment in Metashape
Finally, we would like to extend our thanks to Karen Joyce of James Cook University. We thank them for posing the question, sharing the data and allowing us to use this consultancy example here on our blog.