Registration methods using the MSA in RiScan methods for dense vegetation

Discuss all Riegl related products here.
Post Reply
dylkr
I have made <0 posts
I have made <0 posts
Posts: 2
Joined: Fri Mar 29, 2019 3:58 pm
Full Name: Karun Reuel Dayal
Company Details: IRSTEA
Company Position Title: PhD student
Country: France
Linkedin Profile: No

Registration methods using the MSA in RiScan methods for dense vegetation

Post by dylkr » Wed Apr 03, 2019 5:51 pm

Hello all,

I have been working with approximately 40 scans(in pairs of vertical and tilted scan positions) for a forest. There were significant challenges in the acquisition of the data due to which the reflectors were only used when feasible. Several scan positions were co-registered using the coarse registration method. The primary objective was to ensure that all the scans were co-registered so as to have a complete dataset. That has been achieved but, however, there is a lot of room for improvement to make the data usable for further analysis.

This brings me to my question. Are there reliable and tested methods of using the MSA solely for the purpose of co-registering vegetation datasets? The challenges in this scenario are different since there are no significant planar features which would allow for the standard process of using plane patch filters (and the newer 'extract plane patches') and final adjustment( the problem is further compounded while improving the co-registration between vertical and inclined positions as there is no ground in the latter). I would appreciate some guidance in this regard and hopefully, initiate a discussion on the parameters meant for co-registering vegetation datasets. I was unable to find studies/posts that discussed this matter in-depth.

Thank you!

tbwester
V.I.P Member
V.I.P Member
Posts: 878
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: Registration methods using the MSA in RiScan methods for dense vegetation

Post by tbwester » Wed Apr 03, 2019 8:32 pm

you really need hard targets in between setups to register the data accurately in this kind of environment.

Sounds like that is not an option now. You may be able to line it up in Cloud compare and bring the data together manually, by eyeballing the alignment using trees and other noticeable overlapping features. You wouldn't want to use any cloud 2 cloud algorithms, just manually bringing them together, IMO.

dhirota
V.I.P Member
V.I.P Member
Posts: 522
Joined: Sun Nov 01, 2009 11:18 pm
Full Name: Dennis Hirota
Company Details: Sam O. Hirota Inc.
Company Position Title: President
Country: USA
Linkedin Profile: Yes
Location: Hawaii, USA

Re: Registration methods using the MSA in RiScan methods for dense vegetation

Post by dhirota » Wed Apr 03, 2019 10:51 pm

If you have access to the RiSCAN Pro license, Riegl issued an updated version a few days ago, Version 2.8.0 that has a MSA Version 2. I believe to have it working at optimum, you need targets.

What version are you using?

We have used our VZ400i, scanning in dense vegetation using Trimble R10 in RTK mode for registration, but not under trees to capture the GNSS signals.
Dennis Hirota, PhD, PE, LPLS
dennishirota@gmail.com

User avatar
joe.3dlm
V.I.P Member
V.I.P Member
Posts: 293
Joined: Fri Nov 30, 2007 9:48 am
Full Name: Joe Beeching
Company Details: GeoSLAM Ltd
Company Position Title: QA Lead
Country: UK
Linkedin Profile: Yes
Location: Nottingham, UK
Contact:

Re: Registration methods using the MSA in RiScan methods for dense vegetation

Post by joe.3dlm » Thu Apr 04, 2019 9:23 am

Hi all,

I've had greast success registering VZ data using the Multi-Station Adjustment in RiSCAN PRO, even in difficult environments. It's the best scan-to-scan registration algorithm I've used, especially with the new Automatic Registration 2.0.

Did you use a Riegl tilt mount to capture the tilted over scans? This would be my recommendation for forest scanning. You can easily create a tilt mount calibration, which defines the rotation/translation between an "upright" scan and a tilted over scan. Then, you don't need to worry about registering the tilted over scans - as they are tied to the upright scans by this calibration.

You can just register the upright scans together and the tilted over scans would be carried with them.

As for settings, I would generally run the "Automatic Registration 2" first, making sure to choose one of the "indoor" scenarios. This will generate the required registration metadata, and if you are lucky, will do a reasonable job of registering some or all of the scans together. And if it fails, you already have the prepared registration metadata ready to manually line up the scans.

Then, choose one scan to be the "reference", which you keep locked. Match the neighbouring scans on to it, working out from the middle in rings (like an archery target) using the "Modify orientation and position" tool: https://www.youtube.com/watch?v=dpHmnYpoy50

Run the MSA on just those scans, making sure you have locked the reference scan. The MSA parameters will have already be left in a usuable state by the Automatic Registration 2 tool, but you will need to adjust the search radius. Because you have already moved your scans close together, use a search radius of 1m, run it, then drop the search radius to 0.3m, and run again. Then, check the registration and lock that "ring" of scans, before moving outwards to the next ring and repeat.

Once finished, you will then need to copy your tie points to the project tie point list from the scanner tie point lists, and then to a "find corresponding points" between the project TPL and the global TPL to global georeference your data.

Let me know your thoughts!

Joe

dylkr
I have made <0 posts
I have made <0 posts
Posts: 2
Joined: Fri Mar 29, 2019 3:58 pm
Full Name: Karun Reuel Dayal
Company Details: IRSTEA
Company Position Title: PhD student
Country: France
Linkedin Profile: No

Re: Registration methods using the MSA in RiScan methods for dense vegetation

Post by dylkr » Tue May 07, 2019 2:54 pm

Hello everyone,

Thanks for all the replies!

I appreciate the fact that targets are essential in a forest environment. Scans were carried out at some positions with targets but the same could not be implemented everywhere.

However, I have been able to solve the problem. Firstly, I had to manually co-register the dataset. There was no other option. The vertical and inclined scans were acquired as independent scans and unfortunately, the tilt mount calibration was not used in the field. Co-registering these datasets was not a significant problem as the features were easy to identify.

Secondly, I used the MSA with the dedicated 'Extract plane patches' to align all the vertical scans first, followed by the inclined scans. Somehow this seemed to produce good results.

It may be worthwhile to understand if there is indeed a difference in the results when this procedure is followed when compared to sequentially aligning the scans.

Thank you once again!

Post Reply

Return to “Riegl”