Hello all,
I’m trying to register a gantry crane that did not have proper target placement across the gantry and the overlapping geometry is not sufficient enough for planes or picked points. Rescanning is not an option as the crane is rarely on shutdown and available.
I’ve read on the forum that the fit pipe command and the resulting cylinders can be used for registration. So far I’ve been able to create subscans each with a pipe and bring them into a cluster, but I do not know how to register the scan together or even if I am doing the fit pipe correctly as it was very time consuming for just 8 pipes across 2 scans.
Any help would be appreciated,
Thanks
Registration in scene using pipe fit
-
- I have made <0 posts
- Posts: 2
- Joined: Tue Jun 08, 2021 9:48 pm
- 2
- Full Name: James Kretz
- Company Details: JNE Consulting
- Company Position Title: Civil Designer
- Country: Canada
- Linkedin Profile: Yes
Registration in scene using pipe fit
You do not have the required permissions to view the files attached to this post.
- jcoco3
- Global Moderator
- Posts: 1724
- Joined: Sun Mar 04, 2012 5:43 pm
- 12
- Full Name: Jonathan Coco
- Company Details: Consultant
- Company Position Title: Owner
- Country: USA
- Linkedin Profile: No
- Has thanked: 70 times
- Been thanked: 157 times
Re: Registration in scene using pipe fit
Hi James, welcome to the forum
From your screen shots that does not seem like a great deal of data for a couple of scans. Were you out in the middle of a field or did you do something else like heavily filter the data, or maybe you are using a clipping box? The reason I ask is that often it can be easier to use buildings and structures in the background of the scan for registration as an alternate to the object(s) of concern. The pipe fit tool is kind-off a last resort for me, and if I was presented with the same problem I would probably just try cloud to cloud first as I think it would likely produce better results. Why?...you ask, because in my experience the pipe fit tool doesn't produce the most consistent results (without using tremendous care) especially from different scans and perspectives. Just try to create the same pipe twice in the same scan and then compare the two pipes and you will see what I mean.
If cloud to cloud didn't work well, I would probably still try picking points first, then pipes absolutely last. Even then it is usually more difficult since from scan to scan you will be creating a very similar constellation of targets as the crane's structure repeats. This will inevitably lead to more work creating anti-correspondences and such to prevent all your scans from ending up in the same pile. I don't even go this route any more much less teach it, as cloud to cloud has for the most part eliminated that hassle.
Hope this helps
From your screen shots that does not seem like a great deal of data for a couple of scans. Were you out in the middle of a field or did you do something else like heavily filter the data, or maybe you are using a clipping box? The reason I ask is that often it can be easier to use buildings and structures in the background of the scan for registration as an alternate to the object(s) of concern. The pipe fit tool is kind-off a last resort for me, and if I was presented with the same problem I would probably just try cloud to cloud first as I think it would likely produce better results. Why?...you ask, because in my experience the pipe fit tool doesn't produce the most consistent results (without using tremendous care) especially from different scans and perspectives. Just try to create the same pipe twice in the same scan and then compare the two pipes and you will see what I mean.
If cloud to cloud didn't work well, I would probably still try picking points first, then pipes absolutely last. Even then it is usually more difficult since from scan to scan you will be creating a very similar constellation of targets as the crane's structure repeats. This will inevitably lead to more work creating anti-correspondences and such to prevent all your scans from ending up in the same pile. I don't even go this route any more much less teach it, as cloud to cloud has for the most part eliminated that hassle.
Hope this helps
-
- I have made <0 posts
- Posts: 2
- Joined: Tue Jun 08, 2021 9:48 pm
- 2
- Full Name: James Kretz
- Company Details: JNE Consulting
- Company Position Title: Civil Designer
- Country: Canada
- Linkedin Profile: Yes
Re: Registration in scene using pipe fit
The original scans contain just the ground and barely catch the face of one building (corugated iron so no planes possible). I filtered the distance down so that the rotation grip point for manually placing the scans was at the center of the scan and not way off in space making it easier to finely place the scans.
Cloud to cloud places the scans ok (3-5cm gaps) but for the as-built tools and analysis I want to perform after this, I need a fairly tight point cloud. I've gone to manually placing the scans using the transformation window but it becomes impossible when the scans coordinate system is not aligned with the scan geometry so you have to adjust 2 rotation parameters at a time. I could continue down this route if there was a conisistent way to properly level the scans?
Also thank you, I've been lurking for 2 years now and have finally made an account
Cloud to cloud places the scans ok (3-5cm gaps) but for the as-built tools and analysis I want to perform after this, I need a fairly tight point cloud. I've gone to manually placing the scans using the transformation window but it becomes impossible when the scans coordinate system is not aligned with the scan geometry so you have to adjust 2 rotation parameters at a time. I could continue down this route if there was a conisistent way to properly level the scans?
Also thank you, I've been lurking for 2 years now and have finally made an account
You do not have the required permissions to view the files attached to this post.
- stevenramsey
- Global Moderator
- Posts: 1937
- Joined: Sun Aug 12, 2007 9:22 pm
- 16
- Full Name: Steven Ramsey
- Company Details: 4DMax
- Company Position Title: Technical Specialist Scanning
- Country: UK
- Skype Name: steven.ramsey
- Linkedin Profile: Yes
- Location: London
- Has thanked: 30 times
- Been thanked: 72 times
- Contact:
Re: Registration in scene using pipe fit
Looks like plenty of data for a C2C, maybe not in Scene. If the pipes work you need at least three in each axis to really make it work
- jcoco3
- Global Moderator
- Posts: 1724
- Joined: Sun Mar 04, 2012 5:43 pm
- 12
- Full Name: Jonathan Coco
- Company Details: Consultant
- Company Position Title: Owner
- Country: USA
- Linkedin Profile: No
- Has thanked: 70 times
- Been thanked: 157 times
Re: Registration in scene using pipe fit
Sorry for my delay. Agree with Steve here. Should be possible for cloud to cloud.
First, your filtering seems appropriate, so scratch my previous thoughts.
Second, there is a way to "re-level" scans in Scene somewhere in the properties window of each scan. From what I can remember you have to do it one at a time to each scan, and it basically just resets the scan tilt back to it's inclinometer reading(right or wrong). I won't say that is the "proper" way to level a scan, because then we have to talk about other things like survey control/elevation data/inclinometer issues/etc, but I think it might help you if your scans get tilted due accidental rotations from your manual manipulations in the correspondence view.
Third, what is your current cloud to cloud strategy? The reason I ask is that not everyone seems to use it the same way, and if expectations are that is should work on the first attempt then you will be greatly disappointed. I have always treated cloud to cloud with an iterative approach refining the settings after each attempt while also giving hints to the software through manual manipulations in correspondence view. If that is you strategy then it is good to "update" the registration so that your hints are incorporated. Probably something you already knew, just thought I would mention it. This may or may not be all that apparent in the newer interface.
After you get the cloud to cloud where it has 3-5cm gaps try to reduce the search radius to 0.3m or less. You can also play around with the sub-sampling, but keep in that less sub-sampling (more-data being used) doesn't always end up with better results, as this depends on object geometry of things are in the scan. I generally don't increase the iterations until I start merging many scans and clusters.
Once again you might have already tried this strategy, just trying to eliminate the obvious.
I love the painted on image comments btw
First, your filtering seems appropriate, so scratch my previous thoughts.
Second, there is a way to "re-level" scans in Scene somewhere in the properties window of each scan. From what I can remember you have to do it one at a time to each scan, and it basically just resets the scan tilt back to it's inclinometer reading(right or wrong). I won't say that is the "proper" way to level a scan, because then we have to talk about other things like survey control/elevation data/inclinometer issues/etc, but I think it might help you if your scans get tilted due accidental rotations from your manual manipulations in the correspondence view.
Third, what is your current cloud to cloud strategy? The reason I ask is that not everyone seems to use it the same way, and if expectations are that is should work on the first attempt then you will be greatly disappointed. I have always treated cloud to cloud with an iterative approach refining the settings after each attempt while also giving hints to the software through manual manipulations in correspondence view. If that is you strategy then it is good to "update" the registration so that your hints are incorporated. Probably something you already knew, just thought I would mention it. This may or may not be all that apparent in the newer interface.
After you get the cloud to cloud where it has 3-5cm gaps try to reduce the search radius to 0.3m or less. You can also play around with the sub-sampling, but keep in that less sub-sampling (more-data being used) doesn't always end up with better results, as this depends on object geometry of things are in the scan. I generally don't increase the iterations until I start merging many scans and clusters.
Once again you might have already tried this strategy, just trying to eliminate the obvious.
I love the painted on image comments btw