Using point cloud in Vectorworks

[Neelambari Shukla]

Hi,

    I am trying to learn point cloud data import into Vectorworks for buildings. Could some one point out to me articles or videos, tutorials for getting started on this.

• Which point cloud format works best with Vectorworks
• Are there any third party plugins needed to accomplish this.
• Can I get point cloud data for few single family homes to practice.

 

I have VSS membership. Does it have any special offering for point cloud data processing. 

thanks

Neelambari

 

[Tom W.]

See this video:

 

 

You can import point clouds in various file formats but E57 is the format I've used.

 

 

[Neelambari Shukla]

Thanks for sharing this and getting me started.

[LarryO]

Good luck with your research on that. I took a week in an abandoned attempt to utilize the technology to capture basic as-built information within staircases for a large renovation. Vectorworks simply choked on the data overload for which it doesn't seem to have a means aggregate similar points into planes, lines or surfaces. Manually processing the imported data was a non-starter. Simple rotation and zooming in 3d space was an exercise in patience.

[Tom W.]

2 hours ago, LarryO said:

it doesn't seem to have a means aggregate similar points into planes, lines or surfaces. Manually processing the imported data was a non-starter.

 

Can you explain what you mean?

 

2 hours ago, LarryO said:

Simple rotation and zooming in 3d space was an exercise in patience.

 

Ditto.

[LarryO]

In the data set that was collected there was a density of points (which I don't have access to anymore) that when looking at them in Vw's 3d drawing space we as humans could visualize the irregularities and similarities of the data representing a stairwell formed from concrete. I could see where there was misalignment from overlapping data scans as the surveyor mapped out the levels of the interior of the space. What we needed out of the data was dimensional data primarily in an orthogonal format to produce a railing system that avoided juts, projections, shifts and voids in the as-built space. The point data needed to be reconfigured into surfaces with edges. With over 10GB of points on a single layer attempting to align to any particular mass of points representing a wall, riser, tread, ceiling in order either create a flat plane or surface to which one could measure from or to was near impossible. The rendering of the view would take 90-120 seconds every time a slight change was need to get the view from one place, angle or magnification to another. I attempted to dumb down the data with an external format convertor and within VWs tool set but the tools and engine is not there in VWs to manipulate and transform such large quantities of data into something the application is suited for.

If I had found a route to success I probably would have a better description of the frustration at the enormous amount of time wasted in attempting to modernize (efficient and reliable) the collection of and accuracy of data defining an as-built construction.

 

The current system reminds me of the telephone game children used to play by passing along a bit of information and laughing at the end at how the sentence had been corrupted by the time it was repeated by the final participant in the circle. If the data collector is someone other than yourself you will not receive the whole picture needed to make all the decisions require to produce your work and even then you may not have captured enough data in one, two or three visits because the outcome of the documents was not predefined and was more of a work in progress/design.

 

The reality is that VWs does not come with hard copy manuals explaining the functionality of its tools and specifics of how to implement each. Something which has been lost in the past 30 years of software production and is not limited to VWs. Even Autocad no longer produces its encyclopedic manuals with quick to scan indexes and ability to insert bookmarks. Search functionality even when properly implemented is next to useless if you don't know the terminology used by the author and if the command/tool descriptions are not complete with actual examples of their implementation. Kind of like learning a language without knowing the meaning of the words to be able to choose those words that are applicable to the context. The result is slow and fraught with unpredictable results.

[Tom W.]

@LarryO I'm no expert but a 10GB file for a survey of a stairwell sounds enormous. I have only worked on two projects with point clouds so far (both internal/external scans of buildings) but the largest file I had was 2.3GB, and only then because (for reasons I'll go into later) I was forced to import the scans with 100% of the points. Normally you'd reduce the number of points you import to leave you with a manageable file.

 

3 hours ago, LarryO said:

I could see where there was misalignment from overlapping data scans as the surveyor mapped out the levels of the interior of the space.

 

Again this isn't right + is perhaps why your file was so huge: because the overlapping/misalignment of the scanning resulted in many more points than you should have had? This sounds like a fundamental fault with the survey rather than something inherent to point clouds or VW.

 

My process has been to import my scans at 100% density into a single file + use this as a master file (a template) from which to create all subsequent files from. So my last project was a single site that comprised three separate buildings + each building had an interior scan + an exterior scan which resulted in six point clouds in total. I imported all six into a single file along with the 3D topo data. The reason why I had to import 100% of the points was because I had to get the surveyor to include a special 'datum point' in each file to allow me to align them all correctly. This IS a fault with VW: I found that I couldn't import a point cloud directly into a georeferenced file without suffering a loss of graphics. I instead have to import them at the User Origin initially then manually move them to their correct geographical location using the specially included 'datum point'. Importing the clouds at anything less than 100% risked excluding the datum point!

 

This resulted in a file that included all my survey data for the whole project, albeit at a size of 2.3GB which is obviously too large to do much with. So then it's a process of using this master file as a template + creating new files from it for the areas of the site I wanted to work on: I'd open it up, save a new version + immediately use 'Isolate Points' to remove all superfluous or extraneous points to leave only those relating to the area I wanted to work on, reducing the file size in the process. The files were still in the 800-900MB region but once the initial modelling had been done the point cloud layer/s could be deleted completed to bring the size back down again. If it wasn't for the georeferencing issue the file sizes wouldn't need to be this large.

 

Working with the point clouds is a bit of an art. There is almost too much information + it can be quite a taxing process translating the complex reality of the building into something clean/straight/flat in VW. I guess it depends how clean/flat/straight the building you're scanning is to begin with (in my case not very). Simply viewing the point cloud in its entirety can be a bit overwhelming + very hard to get anything from it. You have to use the clip cube, working planes + saved views to create real time sections through the architecture to enable you to see what's going on + translate it into VW geometry. You are just tracing over a 3D representation. It can be slow + laborious. But then not as slow + laborious as trying to survey it all by hand + create an accurate model that way. The scan will pick up anything + everything so rooms full of furniture + exteriors covered with vegetation are going to make for a much harder to read point cloud. But it sounds like your issue was to do with the fact that you had multiple misaligned scans so any attempt to cut a section through the architecture resulted in a confusing mess rather than relatively clean, easy to apprehend lines...

 

@Luis M Ruiz's video describes the process very well. What is not explained anywhere is why if you import a laser scan into a georeferenced file you end up with a horribly stripy point cloud with half the points missing. It took me a week to figure out what was going on + if it wasn't for my very helpful surveyor (who actually downloaded a trial version of VW in order to see the problem for himself) I think I'd still be in the dark now. You shouldn't have to include a common datum point in each file in order to manually geolocate it + I hope that VW will address this issue soon but I haven't received any info as yet to suggest it's being looked at. The whole point of georeferencing (for me) is that I can import multiple surveys + they all align correctly with one another automatically. To have to manually locate six point clouds is a bit of a flaw + especially so when it's only possible when you have a savvy + helpful surveyor willing to include a special point in each file.

[Jeff Prince]

@LarryOthe situation you describe was a failure of the surveyor and/or the person who ordered the survey.

 

When you request a point cloud, you should specify the surveyor is responsible for combining the data, cleaning it, providing a source model and a decimated version.  This gives you the ability to use the decimated model to orientate in vectorworks and then easily swap it out with the high resolution version (if/as required).  If you want to edit a point cloud yourself, Vectorworks is not the correct tool for the task.

 

I’m with you on software manuals and the failings of companies to provide usable references like help menus.  Most of the time, the writing style and format of the Vectorworks help menu seems to be designed to create confusion.  I think it is part of the marketing strategy for their paid training, which is fine of course… they don’t charge much for the software and are slow to fix problems so they have to make a living somehow 🙂

 

[LarryO]

8 hours ago, jeff prince said:

I’m with you on software manuals

 

There was an art to indexing a technical document. A casual review of the index could show the quality of a manual's content and that of the index creator. The explosion of small independents in the publishing industry seems to have killed off that profession with efficiency cuts. A table of contents simply cannot compete with a well developed index.

[Anders Blomberg]

On 2/27/2022 at 10:40 AM, Tom W. said:

To have to manually locate six point clouds is a bit of a flaw + especially so when it's only possible when you have a savvy + helpful surveyor willing to include a special point in each file.

 

@Tom W. Just ran into this issue. Would you have any ideas of how to correctly position the point cloud without the helpful surveyor? I tried to manually drag/match the full (but not geolocated) cloud to the striped (but correctly geolocated) cloud but it really isn't a very good method 🙂

[Tom W.]

25 minutes ago, Anders Blomberg said:

I tried to manually drag/match the full (but not geolocated) cloud to the striped (but correctly geolocated) cloud but it really isn't a very good method

 

So did this method work out in the end? I guess it relies on identifying a specific point in the first cloud + using it to move that cloud to the exact same point in the second cloud...? No not a great solution. Sorry my experience is very limited I have no idea what other options there are + what other people do. What'd be great is if someone from VW could reassure us that they're aware of the issue + suggest if/when it might be resolved. @Luis M Ruiz? Seems crazy that you can't import a point cloud straight into a georeferenced file.

[Anders Blomberg]

Yes, I created 2 separate shuttle files with correct/incorrect positions respectively. Linked them into my main project file och manually dragged the cloud with the incorrect position to match the cloud in the correct position. I believe I got it within a few cm.

[Ian M.]

This conversation explains why there is a software program called MeshLab (https://www.meshlab.net).  I have downloaded it, but stopped in my tracks when I saw all the tools and options.  I feel confident that it can solve the issues that have been discussed above, but I definitely don't have any of the necessary skills yet.

 

I would love to hear from the Vectorworks community which point cloud file format they like to use?  There are so many options that I hardly know where to start on deciding which format to use.

 

I really like https://3dscannerapp.com

[RJ Johnson]

I came across this string, as I am for the first time delivering PointClouds to a Vectorworks client (also it is their first time using a pointcloud), we typically deliver to BricsCAD, Rhino, Bentley and Autodesk users.

 

So coming from the side of Supplying PointClouds to all industries from Architects to Civil Engineers to Mechanical and all else.  First let me say.  To have PointClouds that are even 100 to 200GB in size is very common in todays era.  Most software can handle these with absolutely no issues.  Of note our largest PointCloud ever was Twenty One Terabytes..  The one I am delivering to the Vectorworks client is 117 Scan Stations, all scanned on Low to Medium resolution, and is 37 GB in raw data alone.  The resultant pts file is 73 GB and 2,910,742,105 points.  This is actually a fairly small project, and was exteriors and topographic only.  When scanning interiors, the pointclouds are substantially larger.

 

On the note of cleaning PointClouds.  It is costly.  Be careful what you ask for.  It is very easy to model a PointCloud that has passing vehicles and people captured in it.  Some simple cleaning is easy  and quick.  But to get a nicely perfectly clean PointCloud is immense effort, that takes time and ties up the computer and software license.  The Cleaning effort can even exceed the cost of the initial PointCloud.

 

With the Scanners today, we are capturing at rates up to 2 million points per second.  A single scan station from one single location can easily exceed 35 million points.

 

Also only request your point clouds in color if absolutely necessary.  Color costs, time, money and requires more powerful computers.   Most elements are actually easier to visualize and see in intensity view.  

 

Yes any true Laser Scanning Operator, should have absolutely no problem delivering you a PointCloud on any Coordinates and rotation you want.  We often orient the PointClouds to the Longest Gridline as requested by the architect.

 

All in all, I'd like to learn more about Vectorworks, and would also be happy to share my experience from the Laserscanning and Pointcloud side.  I personally have been scanning since 2004.  Also I'd be interested to speak with people from Vectorworks and create some more knowledge for PointClouds inside of Vectorworks, as there is not much openly available from my research.

 

Sincerely

RJ.

Hope to talk to many soon.

[mhersland]

Hi All 

I am opening this thread, as I hope anyone can help me a bit on point clouds. 

I am quite new to point clouds. Been using smaller scans earlier, but ran into a larger project now. 

 

We had a building scanned from the surveyor. 

The building is about 5.600m2. 

 

The scan exceeded the maximum amount of points Vectorworks could import. 

Therefore I revised 5 files, containing a part of the building. 

Each scan is about 95.860.766 points. 

 

However the resolution seems very rough, when zooming in. 

I made a video of a small column to show it. 

 

Compared to the great tutorial earlier in this thread, my resolution seems way to low to actually draw from. 

 

Apart from that, one part fills 50gb in my RAM, so my computer is really struggling, handling the file. Especially compared to the performance they get in the tutorial. 

I received the point clous as LAZ files. 

 

Do any of you have an idea, if this is what to expect, or if there is a problem with the scan? 

Best from Mathias

 

[Tom W.]

7 minutes ago, mhersland said:

Hi All 

I am opening this thread, as I hope anyone can help me a bit on point clouds. 

I am quite new to point clouds. Been using smaller scans earlier, but ran into a larger project now. 

 

We had a building scanned from the surveyor. 

The building is about 5.600m2. 

 

The scan exceeded the maximum amount of points Vectorworks could import. 

Therefore I revised 5 files, containing a part of the building. 

Each scan is about 95.860.766 points. 

 

However the resolution seems very rough, when zooming in. 

I made a video of a small column to show it. 

 

Compared to the great tutorial earlier in this thread, my resolution seems way to low to actually draw from. 

 

Apart from that, one part fills 50gb in my RAM, so my computer is really struggling, handling the file. Especially compared to the performance they get in the tutorial. 

I received the point clous as LAZ files. 

 

Do any of you have an idea, if this is what to expect, or if there is a problem with the scan? 

Best from Mathias

 

 

Are you importing the point cloud into a georeferenced file by any chance? I have found that unless a point cloud is imported into a file where the User Origin + Internal Origin are coincident I get linear banding caused by missing points, similar to how your point cloud looks.

 

Re the resolution, what is your 'Details' setting set to in Shaded Options? If not 'Very High' try setting it to that + see if it makes a difference.

[mhersland]

Hi @Tom W.

Thanks for quick reply! 

It may be. It is placed 651.120 meters  away from origin. 

 

I tried to "center to VW origin upon import. 

It helped a lot on resolution. 

 

 

It also seems to help on RAM usage, is that your experience as well? 

 

[Tom W.]

That's good news!

 

The points still look very large though. What happens if you change the projection from perspective to orthogonal...?

 

I don't think I've had any issues with large RAM per se, but have found (as you might expect) that a very large point cloud (or collection of point clouds if there was more than one scan) results in a very large + unwieldy file. So what I have done is import everything into a single 'master' file initially, then immediately save a new version + use 'Isolate Points' to crop the scan to the area I want to work on, reducing the size of the scan to result in something more workable. In this way you can divide the point cloud into sections by keeping going back to the 'master' file + build your model bit by bit without needing to have the whole of the scan present at any one time.

 

But this isn't necessarily going to help the RAM problem you're talking about.