Completeness/Coverage of point cloud

[we082218]

Hi Guys, first of all thank you Daniel for the great software! I am attempting to measure the quality of a point cloud, one of the metrics that I am trying to measure is the completeness/coverage of the pointcloud, a researcher in one journal article did this by splitting the point cloud into equal regions, then he divided the number of regions that contained ANY point by the TOTAL number of regions to get the percentage of coverage, I am not sure how I can do this in CC, and please if anyone has any other ideas to measure the coverage of the pointcloud please enlighten me :) Thanks in advance

[daniel]

I think the Rasterize tool should help you. You can project the cloud in a regular grid. Then if you export this grid as a cloud (leaving the empty cells empty) you'll get the number of non empty cells (this will be the output cloud size) and then compare it to the theoretical grid size.

And you can also output the cell density as a scalar field (this way you'll be able to get more statistics - you can compute the mean density, the standard deviation, etc.).

[we082218]

I think the Rasterize tool should help you. You can project the cloud in a regular grid. Then if you export this grid as a cloud (leaving the empty cells empty) you'll get the number of non empty cells (this will be the output cloud size) and then compare it to the theoretical grid size.

And you can also output the cell density as a scalar field (this way you'll be able to get more statistics - you can compute the mean density, the standard deviation, etc.).

Thank you Daniel ! I will be trying that asap ! and I will let u know

[we082218]

I think the Rasterize tool should help you. You can project the cloud in a regular grid. Then if you export this grid as a cloud (leaving the empty cells empty) you'll get the number of non empty cells (this will be the output cloud size) and then compare it to the theoretical grid size.

And you can also output the cell density as a scalar field (this way you'll be able to get more statistics - you can compute the mean density, the standard deviation, etc.).

Hi Daniel !

I used the Rasterize tool to create a grid, and exported the cloud (used the population per-cell as SF). So now I have number of points in this points cloud (which is theoretically the number of non-empty cells). However, I don't know how to get the total number of cells, I mean how do you get the total number of cells (theoretical grid size) ?

[daniel]

Well, as you have the grid size (it's displayed below the grid step in the Rasterize tool), you simply have to multiply the width by the height :D

[zheng]

I think the Rasterize tool should help you. You can project the cloud in a regular grid. Then if you export this grid as a cloud (leaving the empty cells empty) you'll get the number of non empty cells (this will be the output cloud size) and then compare it to the theoretical grid size.

And you can also output the cell density as a scalar field (this way you'll be able to get more statistics - you can compute the mean density, the standard deviation, etc.).

Hi Daniel,

Is 'Non-empty cells/theoretical grid size' the same value as shown in Rasterize-Volume-Non empty cells?
Thank you.

Zheng

[daniel]

Hum... I can't remember ;)

Just to be sure you should export the cloud (with the 'empty cells' option at 'leave empty') and check that the number of resulting points against the size of the grid (W * H). Just to be sure ;)

[Shakhaa]

Hi. I am working on something similar to calculate the coverage of a point cloud. Could you please name the paper or the journal you are referring to ?

[user3]

Hello.
I have a reference mesh that is 100% complete. On the other hand I've a point cloud (with gaps) who's completeness I should assess, regarding to the reference mesh. Is it possible to do that in 3D space and how?
Thank you.

[daniel]

Not really (I mean not in 'full 3D'). Unless it is acceptable to sample point on the mesh, and then compute the distances between the resulting cloud and the partial cloud (as reference) so as to detect/quantify the missing parts?