I am doing a comparison of a Limestone cave that has two scans prior and post construction work. I am using CC to estimate any changes in the cave surface which is all ok apart from the fact that the software detects changes where there are points present that were not in the original (reference) dataset. In other words, due to differences in scanner position in the two surveys, some areas of shadow in the initial visit have point in the follow up second visit.
Is there a way of making the coverage of clouds similar so that the only changes between the pointcloud distances are due to real features as opposed to mapping out the shadows in the scan data?
Many thanks for any advice.
Kind regards
Nick
Comparing clouds from cave survey - mask unwanted points
Re: Comparing clouds from cave survey - mask unwanted points
Hi,
This is an interesting question.
By default when you compare two clouds you have to make sure the extents of the reference cloud is bigger/wider than the compared cloud. One can always segment the compared cloud so as to remove the non-overlapping parts (with the scissors icon). But during my PhD I developed a mechanism to use the sensor position information to do this automatically. But at that time importing the sensor pose and position was very cumbersome (as no manufacturer was sharing their data format specifications). So I eventually removed this from CloudCompare. Now that we can import this information more easily (from FARO or PTX files) I could re-enable it.
What file format/sensor are you using?
This is an interesting question.
By default when you compare two clouds you have to make sure the extents of the reference cloud is bigger/wider than the compared cloud. One can always segment the compared cloud so as to remove the non-overlapping parts (with the scissors icon). But during my PhD I developed a mechanism to use the sensor position information to do this automatically. But at that time importing the sensor pose and position was very cumbersome (as no manufacturer was sharing their data format specifications). So I eventually removed this from CloudCompare. Now that we can import this information more easily (from FARO or PTX files) I could re-enable it.
What file format/sensor are you using?
Daniel, CloudCompare admin
Re: Comparing clouds from cave survey - mask unwanted points
Hi Daniel,
Thanks for your thoughtful reply. It's tricky since some of the scanner positions for the repeat survey were higher up and have less shadowing than the reference dataset.
The data was acquired with a Faro Focus X130 and processed in Topcon Scanmaster. So we can export registered scans as .ptx, .e57 etc.
Kind regards,
Nick
Thanks for your thoughtful reply. It's tricky since some of the scanner positions for the repeat survey were higher up and have less shadowing than the reference dataset.
The data was acquired with a Faro Focus X130 and processed in Topcon Scanmaster. So we can export registered scans as .ptx, .e57 etc.
Kind regards,
Nick
Re: Comparing clouds from cave survey - mask unwanted points
Hi Daniel,
I was wondering if you have had any more chances to think about this problem and facilitate the removal of non-overlapping points by using the scanner positional info?
Kind regards,
Nick
I was wondering if you have had any more chances to think about this problem and facilitate the removal of non-overlapping points by using the scanner positional info?
Kind regards,
Nick
Re: Comparing clouds from cave survey - mask unwanted points
Yep, but it took me some time to understand how the FARO scans were organized (as we need the sensor angular steps to do this properly).
Hopefully I did things right. You'll find a custom version here:
http://www.cloudcompare.org/release/Clo ... 4_FARO.zip
Before comparing your clouds, you'll have to do something on each 'reference' cloud:
- expand the tree under the cloud
- highlight the 'sensor' object (Ground Based Lidar Scanner)
- then go to 'Edit > Sensors > Ground-Based Laser > Show depth buffer'
You should see a depth map corresponding to what the scanner has seen during the acquisition:
Once this is done you can compute the distances. The points that are 'hidden' (i.e. farther than what the scanner has seen) or 'out of field of view' will be ignored. You will be able to hide them afterwards by unchecking the 'show NaN/out of range values in grey' checkbox in the 'Parameters' tab of the scalar field section of the (compared) cloud properties.
I'd be very interested to get some feedback on this as it's quite experimental and I don't have FARO data on my side that allow me to test this properly. If it works I'll add an option to the 'Distances computation' dialog so as to compute the depth map automatically.
Hopefully I did things right. You'll find a custom version here:
http://www.cloudcompare.org/release/Clo ... 4_FARO.zip
Before comparing your clouds, you'll have to do something on each 'reference' cloud:
- expand the tree under the cloud
- highlight the 'sensor' object (Ground Based Lidar Scanner)
- then go to 'Edit > Sensors > Ground-Based Laser > Show depth buffer'
You should see a depth map corresponding to what the scanner has seen during the acquisition:
- cc_depth_map.jpg (52.56 KiB) Viewed 6850 times
I'd be very interested to get some feedback on this as it's quite experimental and I don't have FARO data on my side that allow me to test this properly. If it works I'll add an option to the 'Distances computation' dialog so as to compute the depth map automatically.
Daniel, CloudCompare admin
Re: Comparing clouds from cave survey - mask unwanted points
I have done a very similar process to this and just wanted to add my two cents.
Unless you can control the environment, and setup in the same scan positions as before I see no real way to avoid the shadows. We have just known that when we look at the data some of what is reading as changed, might just be shadows and you need to be mindful of it.
In my head, it is possible to do this automatically but it is cumbersome. I call it Negative scan data. In theory, if you took a scanner position, and its points, you could build a kind of reverse cloud that knows that there is no scan data here, the view was blocked. Sky would be good, there was no obstructions, but starting from the other side of the wall to X there is no scan data.
Image drawing a line from scanner origin to every point. Do this for two scanners, when these lines are cross each other (or are close) you know the scan data was captured in both scanners. If the lines do not cross, you know that's where your shadows are. Carry this information across multiple scanners and you can determine roughly how many times scan data was captured and recaptured from one scan to the next, and where your shadows are.
I will draw diagrams if it would help get this idea across.
Lastly Daniel if you need Faro scanner data its possible I can get you sample sets, just ask what your looking for and if we have it I can send it your way.
Unless you can control the environment, and setup in the same scan positions as before I see no real way to avoid the shadows. We have just known that when we look at the data some of what is reading as changed, might just be shadows and you need to be mindful of it.
In my head, it is possible to do this automatically but it is cumbersome. I call it Negative scan data. In theory, if you took a scanner position, and its points, you could build a kind of reverse cloud that knows that there is no scan data here, the view was blocked. Sky would be good, there was no obstructions, but starting from the other side of the wall to X there is no scan data.
Image drawing a line from scanner origin to every point. Do this for two scanners, when these lines are cross each other (or are close) you know the scan data was captured in both scanners. If the lines do not cross, you know that's where your shadows are. Carry this information across multiple scanners and you can determine roughly how many times scan data was captured and recaptured from one scan to the next, and where your shadows are.
I will draw diagrams if it would help get this idea across.
Lastly Daniel if you need Faro scanner data its possible I can get you sample sets, just ask what your looking for and if we have it I can send it your way.
Re: Comparing clouds from cave survey - mask unwanted points
Well, this is more or less what we do with the depth map and sensor position of the reference/previous cloud.
When projecting any 3D point in the scanner point of view (and for each viewpoint if you have several scans) you can tell:
- that it is out of range (if it's very far and you know the maximum range of the scanner)
- that it was out of field of view (if it falls outside of the depth map 'rectangle')
- and more interestingly you can flag the points that fall 'behind' the depth map because you don't have any reliable information on what happened behind the disappearing wall (is it that the wall collapsed an revealed something else, or has the wall simply be shifted?). From this point you need the human intelligence ;).
And I'd be interested in before/after scans of the same area to tweak this system in CloudCompare (it was working during my PhD but the scans were not as dense as now and they were not 360°).
When projecting any 3D point in the scanner point of view (and for each viewpoint if you have several scans) you can tell:
- that it is out of range (if it's very far and you know the maximum range of the scanner)
- that it was out of field of view (if it falls outside of the depth map 'rectangle')
- and more interestingly you can flag the points that fall 'behind' the depth map because you don't have any reliable information on what happened behind the disappearing wall (is it that the wall collapsed an revealed something else, or has the wall simply be shifted?). From this point you need the human intelligence ;).
- cc_tls_visibility_issue.jpg (62.35 KiB) Viewed 6833 times
Daniel, CloudCompare admin
Re: Comparing clouds from cave survey - mask unwanted points
That image helps paint the picture. There is no good fix other then human eye. The only fix to remove the human element is to have perfect scan data with no shadows.
Daniel, I have Faro scan data that I can share with you of an office taken months apart. The walls didnt change, but the furniture was moved about, new furniture added etc. Would you like FLS files, E57, BIN?
Daniel, I have Faro scan data that I can share with you of an office taken months apart. The walls didnt change, but the furniture was moved about, new furniture added etc. Would you like FLS files, E57, BIN?
Re: Comparing clouds from cave survey - mask unwanted points
Cool!
If you can send me the FARO project it would be perfect.
If you can send me the FARO project it would be perfect.
Daniel, CloudCompare admin
Re: Comparing clouds from cave survey - mask unwanted points
Daniel,
I can send you two scans from a cave before and after some major changes to the rock faces. Would you like raw Faro scans (i.e. unregistered), or shall I register them and send you the data in something like .e57 format?
Best,
Nick
I can send you two scans from a cave before and after some major changes to the rock faces. Would you like raw Faro scans (i.e. unregistered), or shall I register them and send you the data in something like .e57 format?
Best,
Nick