Posted by Andrew Gaiennie on Jun 1, 2011 11:28 am

How hard is it to match point clouds so I don't need targets on a jobsite? And where can I go to learn about using scanners?

Re: Scanning

Posted by Rusty Chain on Jun 1, 2011 11:58 am

Weren't you pretty set on starting a business predicated on providing scanning data as a consultant to engineering and surveying companies just a few weeks back?  If you were prepared to set yourself up as a consultant, I would think that you already have these answers.

If you don't have the answers, I'm sure that Falk or Wimbush (does he ever make an appearance here? Is he still in Africa?) can answer them well.

Re: Scanning

Posted by mwimbush on Jun 1, 2011 12:12 pm

Still in Africa

Re: Scanning

Posted by MLB on Jun 1, 2011 2:53 pm

Andrew- I think what you are referring to is called "Mesh Creation". All of the vendors have training available on the software that performs this operation. Trimble had one on line, but I haven't checked it recently. Perhaps you had best wait until Professor Falk weighs in.

Re: Scanning

Posted by RADU on Jun 1, 2011 7:32 pm

I watched a demo in 2004 in Oregon conference where traversed around a building joining clouds from common points/areas and no survey control. Simply a software issue.

As Micheal B said it called meshing.


Re: Scanning

Posted by Mike Falk on Jun 2, 2011 7:40 am

Cloud to Cloud registration is the term.

Meshing is a practice that is performed after registration. Meshing is putting a Mesh, or surface (TIN), over a point cloud

Are you having difficulty at your current employer? You may want to think about your previous choice "B"

Laser Scanning is a part of our profession that requires SERIOUS capital and time committment to be done well.

Re: Scanning

Posted by Mark Mayer on Jun 2, 2011 10:58 am

Andrew Gaiennie:
How hard is it to match point clouds so I don't need targets on a jobsite? And where can I go to learn about using scanners?
While it may be possible to match cloud to cloud without targets I expect your comfort level would shoot up with targets. I, too have seen the salesman's demo with the cloud to cloud matching. I have not seen the salesman with a license to protect do so. Perhaps if the clouds contain enough sharply defined surfaces it could be done.

As far as where to learn about scanning - that would be mostly in the hands of the vendors.   

Re: Scanning

Posted by Mike Falk on Jun 2, 2011 11:54 am

Mobile laser scanning is an example of cloud to cloud registration without using targets. 

Many of the high end mobile scanning units include multiple scanner heads. As the vehicle moves each scanner head gathers data in a cloud individual to that scanner head. These multiple point clouds must be registered, and typically, cloud to cloud.

This registered point cloud can then be registered, or ground truthed, to project control.

I might suggest attending multiple vendors demonstrations, as each will put their own spin on their process. A vendor's number one goal is to sell equipment and software.

The better a company understands error propagation and process control, the more effective they are at NOT  using targets.

Re: Scanning

Posted by MLB on Jun 2, 2011 2:47 pm

As a tidbit, the recommended (preferred) method for registering multiple point clouds in Trimble Realworks is , set the scanner on a point with a known position and backsight another point with a known position. Whoda thunkit?

Re: Scanning

Posted by Mike Falk on Jun 2, 2011 5:30 pm

The style of "set the scanner on a point with a known position and backsight another point with a known position " registration is one of the least rigorous, weakest, and  lowest quality ways for registering multiple point clouds.

Some vendors use this style of traversing to help some Land Surveyors to better under what is going on.

Re: Scanning

Posted by MLB on Jun 3, 2011 12:10 pm

Just saying, here is an excerpt from Trimble Realworks documentation:

Cloud-, target-, or known-point-based methodologies;
analysis, quality control, and reporting In RealWorks Survey, users can easily register scans together using one of several different methods: cloud-based,
target-based, or via geo-referencing. For cloud-based registration, users pick the same points in two different scans with the option of refining the registration in a second stage. In target-based registration (when appropriate), users benefit from fully automatic station-to-station
registration, as well as quality control and reporting.

Targetbased registration applies to traverse/station setup and free stationing/resection, as well as un-leveled stationing.
For registration via the geo-referencing tool, users can assign known coordinates to various points in all scans to be registered. For example, the points can be target centers, known points, or survey points from other instruments. To further enhance the registration process, the
Target Analyzer tool can be used to do one or all of the following: check if there are enough targets or survey points,modify or delete those
that are incorrectly fitted, create additional targets in the point cloud where such a target is identified visually as having been
scanned. For results reporting, use the Entity-Based Registration Report tool to generate a full quality control report
in .rtf format.

I would suppose it depends on how important it is for the user to know exactly where his rigorous quality point clouds are actually located in the world.

Re: Scanning

Posted by Sprint Car on Jun 3, 2011 2:11 pm

We are using the Faro scanner, not a surveyor friendly product.  You can set it up on known coords and scan away, or you can do more of a resection scan.  Either way seems to work pretty good, however it is all about the control.  If you are side shooting all of your control I would say, not the best way to do it., tried it and did not like the results.  We have had very good results by setting and direct shooting control for distance and coordinate values.  If you are putting things on a coordinate system I am not sure exactly how you would do it without a target of some sort.

I have also had extremely good results simply setting up and scanning and not having any targets at all. The software can find enough to fit the scans, however I can see a problem if you are scanning a pile of something for volumes and not having enough to fit the scans.  Again, not sure how you would do that without targets of some sort.

So, I would think depending on environment, it would be hard to fit clouds together without some kind of targets.  Targets being a vague word.

I have not had any experience with the moble scanning, so I will have to take Mr. Falk's word on how it works.  Although, I did have a guy that knows all about moblie scanning tell me that they process the data to survey grade, any comments?

Re: Scanning

Posted by A F on Jan 2, 2012 11:25 pm

Hey Guys,

This is pretty straightforward: There is not one-size-fits-all workflow.
Targets are mandatory on DOT jobs. There's a reason for this: when you do not have enough matching surfaces between two scans, you will experience a lack of strength in the Cloud2Cloud registration results. Therefore, Cloud2Cloud is not the silver bullet. Keep in mind: scanning is surveying in 3D. Just as you should always have both a backsight AND a foresight when performing high precision jobs, you should utilize targets as frequently as possible when scanning. The targets form the framework for the scans and enable rapid positioning and registration of each scan while the Cloud2Cloud enables adjustment or 'tightening down' of the scans.

I get asked all the time "how do I get rid of those targets, they slow me down?". The answer is: you don't! The same principle that applies in surveying "check it twice, do it right" applies in scanning: use redundant positioning methods. If you're using a total-station method for the scanner (leveling and backsighting) then use two backsights! If you're using a resection method use an extra target (not 3, use 4). Likewise for Cloud2Cloud: If you want to use common planar surfaces, take an extra scan at junction points between two critical scans, this will lock those two together.

Yes, salesmen do not have to sign for a project! Remember that. Make sure that you see more than a pretty picture. Humans' eyes are deceptive: if it 'looks' right then our minds believe it is right. This is total BS. Its right if its right. You can prove a point cloud with traditional means and this is exactly what you should do. If you want to have a demonstration done by a scanner salesman, prepare a test field that you have measured with levels and TS, don't use RTK, its not good enough. Then give the salesman 4 control points to use and expect a georeferenced point cloud at the end. Once that's completed be sure they are "satisfied" that their data is registered properly. If it is, then give them your test points and have them pull in the points and 'prove' it to you. Don't let them walk away and 'just adjust' or elsewise. Again: if its good, it is good.

A final word on Mobile Scanning: a correctly calibrated mobile scanner does not need to be adjusted via a cloud2cloud algorithm. The two (or more) lidar units should be precisely tuned such that together they produce a homogenous point cloud. The error from mobile is introduced through integration and GPS errors. If all work is performed correctly, only the GPS error should need to be adjusted.

Re: Scanning

Posted by mcasey on Oct 11, 2012 8:11 am

It is not much tough. You can learn this. Visit the link provided below to get more information and get all your query clarified by the expert.

lidar scanning