Cyclone Smart Align

Discuss all Cyclone related issues here.
Chris06
I have made 30-40 posts
I have made 30-40 posts
Posts: 35
Joined: Thu Apr 09, 2009 2:19 am
Full Name: Chris
Company Details: Not Applicable
Company Position Title: Not Applicable

Cyclone Smart Align

Post by Chris06 » Fri May 19, 2017 5:32 am

Hi All,

I'm Wondering if anyone has used the smart align function in Cyclone and if they've had much success. We recently (finally) updated to 9.1.5 and I was keen to try out the new smart align tool but I've been very disappointed with the testing.

We had the seemingly ideal project to test it - multi storey, stripped building which we needed to scan but after playing around with the smart align tool I've been very disappointed. I was at Hexagon last year when the idea for the smart align tool was suggested and I had very high hopes but it seems like it has been poorly executed. I can list a number of specific reasons but I thought I'd just ask the forum to see if maybe I had too high expectations or if I'm doing something wrong.

User avatar
pburrows145
V.I.P Member
V.I.P Member
Posts: 1382
Joined: Wed Sep 12, 2007 9:16 am
Full Name: Paul Burrows
Company Details: Leica Geosystems UK Limited
Company Position Title: EMEA Business Manager HDS
Country: UK
Skype Name: pburrows145
Linkedin Profile: Yes
Location: UK

Re: Cyclone Smart Align

Post by pburrows145 » Fri May 19, 2017 8:41 am

Hi Chris,

Where are you based?

Thanks,
Paul

steves01x
I have made 80-90 posts
I have made 80-90 posts
Posts: 83
Joined: Tue Jul 21, 2015 3:58 pm
Full Name: Steve Smith
Company Details: IIC
Company Position Title: Surveyor
Country: Scotland
Linkedin Profile: Yes

Re: Cyclone Smart Align

Post by steves01x » Fri May 19, 2017 9:37 am

Since the day it came out i have used it on every single job!

Some jobs have still needed a day sitting sorting it out but they have been close to 150 scans and less than ideal sites but it still saves loads of time.

Is it failing for you or just putting loads of scans in the ungrouped section?

User avatar
pburrows145
V.I.P Member
V.I.P Member
Posts: 1382
Joined: Wed Sep 12, 2007 9:16 am
Full Name: Paul Burrows
Company Details: Leica Geosystems UK Limited
Company Position Title: EMEA Business Manager HDS
Country: UK
Skype Name: pburrows145
Linkedin Profile: Yes
Location: UK

Re: Cyclone Smart Align

Post by pburrows145 » Fri May 19, 2017 2:42 pm

steves01x wrote:Since the day it came out i have used it on every single job!

Some jobs have still needed a day sitting sorting it out but they have been close to 150 scans and less than ideal sites but it still saves loads of time.

Is it failing for you or just putting loads of scans in the ungrouped section?
Great to hear you're having the success we anticipate from this feature Steve.

The key is to have as much overlap as possible, try the different methods (time-based, manual placement with a site-map etc) and it does also help if you've taken images also (believe or not!).

If you're having difficulty, it might be worth investing in some webex training from your local provider.

See the video here for how it works:


youtu.be/

tbwester
V.I.P Member
V.I.P Member
Posts: 763
Joined: Fri Apr 20, 2012 1:49 pm
Full Name: Thad Wester
Company Details: WeWork
Company Position Title: Head of Reality Capture
Country: USA
Linkedin Profile: Yes
Location: SC/NYC

Re: Cyclone Smart Align

Post by tbwester » Fri May 19, 2017 4:21 pm

we cant get it to work either... even scans right next to each other with probably 80% overlap seem to fail.

Love Cyclone, but one thing it lacks is some automation.

User avatar
Phil Marsh
Administrator
Administrator
Posts: 2008
Joined: Sun Aug 12, 2007 2:28 pm
Full Name: Phil Marsh
Company Details: Laser Scanning Forum Ltd
Company Position Title: Technical Director
Country: UK
Skype Name: phil.h.marsh
Linkedin Profile: Yes
Location: Barnsley, UK

Re: Cyclone Smart Align

Post by Phil Marsh » Fri May 19, 2017 5:43 pm

tbwester wrote:we cant get it to work either... even scans right next to each other with probably 80% overlap seem to fail.

Love Cyclone, but one thing it lacks is some automation.
Can you send me some sample data?

I'll either explain why its not working as it should or show you how to get this work. ;)

Chris06
I have made 30-40 posts
I have made 30-40 posts
Posts: 35
Joined: Thu Apr 09, 2009 2:19 am
Full Name: Chris
Company Details: Not Applicable
Company Position Title: Not Applicable

Re: Cyclone Smart Align

Post by Chris06 » Sun May 21, 2017 11:28 pm

Well it's good to see it's useful for some!

I suppose my main issues with it is that it doesn't seem to have improved the ability of auto align to get a match, and that it hasn't been executed particularly well. Here's a few things I've noticed...

- The actual spatial relationship of how you graphically move the scans around on the editor seems to have no bearing on how successful the auto align is. I quickly worked out it wasn't worth the time to move the scans around based on field notes - my tip here would be to write in your field notes which scans should cloud-to-cloud together and ensure there is a link between them; don't bother wasting your time actually moving them to the correct relative location (unless it helps to draw those links) - the video above even shows that when the user gives up moving them around and just leaves them in their original import location and creates the link by time.

- Just like auto-align has always been, the algorithm is still pretty flaky. For example, we did a test using the main area of our office which is one open plan room, about 12x12m with large windows on 3 sides. I did 4 scans, two of which were only a few metres apart, and none of them were auto aligned. This is probably due to the amount of data collected outside the windows which is not regular and easy for the matching algorithm to spot, but I was still hoping for something better.

- My previous workflow with smart-align was to let cyclone have a go at aligning the whole job, and then once it had got as many matches as possible I would go through and use visual alignment to manually join any groups that it had created or manually add constraints to all the un-grouped scans. The problem with smart align is that if you choose, say, scans 1 &2 and link them together (because the link is obvious and should be easy) and that fails then scan 1 will not be grouped with anything. If I use the old method, there was often a chance that scan 1 might have ended up with a join to scan 3 or 4 or 25 and thus would still be grouped. I'm aware I could draw links between scan 1 to 3 or 4 or 25 if I think it's possible, but this is now a manual process that takes my time. I'd much rather prefer the old method of clicking import before I go to bed, let cyclone play all night if it wants trying to figure out possible matches, and when I wake up I can sort out what it couldn't do. Ok, so maybe 1 in every 5 or 10 auto-aligns is completely wrong, but I feel like I can find them and fix them faster than I can set up all the possible combinations that could occur using the smart align editor.

Which brings me to my thoughts about the usability of the smart align tool...

- First thing I noticed was that when I tried to use a portait-oriented reference image, many of the scans ended up outside the window and I had no way of finding where they had gone and how I could pan to these scans.
- When I clicked the load image button, the now active window popped up behind the main editor window, and then I could continue to click the load image button as much as I liked, opening as many more file explorer windows as I liked, even though the first was still open and waiting... That's just poor programming.
- There's no easy way to just group scans so that they can be tested against every other scan in the group. For example, I scan a multi-level building where there's a strong possibility that every scan on each level can link with every other scan on that level - all I want to do is group all the scans on level 1 so that they are tested against every other scan on level 1 and not on any other level. There's no easy way to do that.
- Creating the link by dropping the scan on top of another scan, then moving it away is painful. Just let us click two scans, or multiple scans, and click "create links".
- If I want to move multiple scans to a new site the only way to select them at once is with a rectangle. What happens when that's possible? I have to do them one by one. Why can't we just use shift/ctrl+click?
- There seems to be no way to only use smart align on a certain number of scans - it always applies to every scan in the registration. What if I had 5 scans outside that I used targets for, then 20 scans inside one building, and 20 in another; I'd like to only select the 20 scans inside the second building and let smart align try and align them, but ignore all the other scans (leaving any constraints that already exist); there seems to be no way to do that.

Anyway... that was just a few issues I found in the first few hours of trying it out. Again, I'm so glad that Leica have listened to their users and have tried to implement this idea - I just wish it had been done better. I'd love to hear if anyone else is finding the same thing, or if I'm just daft and I am doing something (everything?) wrong...

P.S. I'm in Sydney, Aus.

cpw006
V.I.P Member
V.I.P Member
Posts: 267
Joined: Tue Nov 20, 2007 3:51 am
Full Name: Chris Widmer
Company Details: LAS
Company Position Title: 3D Scanning Manager
Country: USA
Linkedin Profile: Yes
Location: Atlanta, Georgia

Re: Cyclone Smart Align

Post by cpw006 » Wed May 24, 2017 6:34 pm

steves01x wrote:
Fri May 19, 2017 9:37 am
Since the day it came out i have used it on every single job!

Some jobs have still needed a day sitting sorting it out but they have been close to 150 scans and less than ideal sites but it still saves loads of time.

Is it failing for you or just putting loads of scans in the ungrouped section?
We have had success with it as well, but it's usually to let it auto-align and then register the groups it creates together.

User avatar
Formula1982
I have made 80-90 posts
I have made 80-90 posts
Posts: 87
Joined: Tue Oct 13, 2015 4:02 pm
Full Name: Stephen Ferrari
Company Details: UTEC StarNet Geomatics
Company Position Title: Technical Lead
Country: United Kingdom

Re: Cyclone Smart Align

Post by Formula1982 » Mon Jun 05, 2017 9:09 am

The only comment I have on this issue, is a pretty self evident one. That if you're using 'automatic' cloud to cloud methods, then the clouds need to have a high degree of overlap. A much higher % overlap than normal scanning in my experience.
There is a tendency when on the job to push further distances between scans, since you're not worrying about whether you can see enough of your existing targets - when in reality you need to do the opposite and add more scans closer together, not stretch them further apart.

jmiller2103
I have made <0 posts
I have made <0 posts
Posts: 8
Joined: Thu May 18, 2017 4:16 am
Full Name: Jay Miller
Company Details: Madigan Surveying
Company Position Title: Surveyor draftsman 3D modeler
Country: Australia
Linkedin Profile: No

Re: Cyclone Smart Align

Post by jmiller2103 » Fri Jun 16, 2017 5:47 am

I believe this to be a great tool, and it does speed up with the registration process. We pick targets in the majority of our scans which definitely improves the success rate of this automation process. With larger data sets, targets are key to helping cyclone understand the positing of the clouds. I believe the scanners will always fall over when it comes the shiny surfaces, windows and very similar data e.g hallways, stairways, railways where even we would find it hard to tell from a photo where it was in relation to another.

My biggest issue is when everything appears to be fine and correct, all of the errors are all below 5mm and your thinking lovely, then once you've created a modelspace you soon realise that it is far from ok, and the registration rubbish, sometimes the fix is simple and only 1 c2c to rectify but when there are multiple in a large dataset and the data is very similar it can be a pain to find the c2c's that are tearing the registration apart. I think I remember the older versions of cylcone registration clearly showing a bad c2c with the general error figures, but now as we keep updating the errors seem to hide themselves in the data.

Registration will always need a little hope with targets, spheres and general notes from the scanning day itself. But to be honest the visual registration is extremely fast aswell so if you need the data to be spot on this is a great tool.

Return to “Leica Cyclone”