can I do change detection with M3C2 algorithm using points clouds with different densities?

[emanuel120]

I am currently doing change detection with the M3C2 cloud compare algorithm for two models from drone images in 2022 and 2023. I have obtained the point clouds and I have noticed that both point clouds have different densities.

when I go to do change detection between the two models, one of the point clouds has more information than the other. Should I resample the data so that the density of both clouds is the same or can I do change detection with point clouds with different densities?

[daniel]

Yes, you can keep the original clouds and use different densities. Just make sure the 'scale' is large enough so that there are enough points in local neighborhoods, even on the low density cloud.

With a more naive algorithm such as the simple C2C distances computation algorithm, you would have to use the high density point cloud as the 'reference' cloud.

[Madeleine]

Hi Daniel,

Considering what you said about taking the cloud with higher density as the reference in C2C, should we do the same with M3C2? I'm wondering what I should put as Cloud #1 and Cloud #2 in the M2C3 distance window if one of my clouds is around 2 times denser than the other. Does it make a difference?

Merci!

Madeleine

[daniel]

With M3C2, that's less an issue, as it copes better with irregular/different densities. It's just that using the higher density cloud as source for the 'core points' (especially without subsampling) will give you a denser result.