Tag: point cloud

Open Source Photogrammetry Workflow with VisualSFM and Meshlab

Luke Johnson

There are a whole lot of paid and open source tools out there for photogrammetry (turning photos into 3D models), and it is easy to get lost. I found a very detailed post by Jesse over at We Did Stuff. While it is a pretty old post, it still has lots of little tips along the way and if you are starting out with some photogrammetry workflows (like in Recap Photo), then this post is worth a read.

 

Here is a short excerpt:

Read the whole post:

http://wedidstuff.heavyimage.com/index.php/2013/07/12/open-source-photogrammetry-workflow/

Oh, and while we are talking about open source point cloud tools, have you had a look at CloudCompare?

 

How to Change The Size of Point Cloud Points in Navisworks

Luke Johnson

Navisworks works really well with Point Clouds, particularly in association with Recap. It will usually create ‘voxels’ – groups of points that you can hide or change colour or use in other Navisworks workflows.

However, sometimes the ‘point size’ seems too fine. To modify this, just open up Navisworks Options to Interface – Display, and change the Primitive size for Points to something that looks better. You can choose any size from 1-9.

pass2b252842529-6795967

Also, there are additional settings under File Readers – ReCap:

recap-2618422

You can set an ‘interactive point size’ here, which is going to override the point display when you are zooming around or navigating the model.

How To Use A Temporary Point Cloud Transformation To Maintain Accuracy in Revit

Luke Johnson

Revit doesn’t like big numbers. There, I said it.

So when dealing with ‘world’ coordinates in a point cloud, sometimes things just don’t work too well. I thought I had this all solved recently by using the DXF, Center-to-Center, Acquire Coordinates workflow. However, I discovered that somewhere along the line, Revit still does break down with the large coordinates. I think this is happening in between Recap and the Revit point cloud rendering engine. I was getting something that looked like this:

As you can see, the shared coordinate system is very large. In this situation, you can’t even move the point cloud into the correct location in Revit, it jumps in large increments when moving. Interestingly, Navisworks and AutoCAD both handle these large coordinates ok – appending the same data does not have the error shown above. So…

How do we fix this and make Revit happy?

Basically, we do a temporary truncation of the source data, get it into Revit, and then reinstate the appropriate coordinate system.

To truncate the data, have a look at your source point cloud information. In my case, I could identify 4 leading digits for the X and Y coordinates that were not significant:

Using EmEditor (which handles large text files very well), and its Vertical Selection feature, I was able to delete the 2781 and 6181 digits from my source data.

In effect, this transformed everything by 278100m and 6121000m. Keep these numbers in mind for future reference…

simplified.png

Ok, with the simplified source data in hand, I followed these steps:

  1. Index a new RCP in Recap using the simplified data
  2. Open surveyor DXF file in AutoCAD and manually Move all the geometry. Move the objects by the values above (278100, 6121000) towards the origin. Save As – a new DWG file with modified coordinates.
  3. Link this modified DWG into Revit, Center-to-Center
  4. Acquire Coordinates from it
  5. Link the Point Cloud RCP By Shared Coordinates
  6. Everything lines up now that the large coordinate shift error has been avoided!
  7. Link in the original DXF and align it with the modified temporary DWG we were using
  8. You may need to temporarily neutralize coordinates (here or here), and…
  9. Now you can Acquire Coordinates from the original DXF and you will have reinstated the ‘world coordinates’, but the Revit point cloud rendering engine is now much happier.

Hope this helps you if you face a similar problem 🙂

Previous post:
What Revit Wants: Using a DXF to Locate a Point Cloud in Revit with Very Large Coordinates

Using a DXF to Locate a Point Cloud in Revit with Very Large Coordinates

Luke Johnson

I thought that most of this was ‘easy’ and solved now, but it was more of a challenge than I expected. I received a ASC file from a survey in XYZRGB format, which looks like this:

asc.png

Those XYZ values are Metres (or Meters if you are in US) in the MGA 94 coordinate system. I also received a DXF file with the same World coordinates, and project related gridlines so I could relate the point cloud to our Revit models.

I tried getting the MGA Shared Coordinates right in Revit, and then linking an RCP or RCS from Recap ‘by Shared Coordinates‘, but I didn’t have much joy.

Here is the workflow that worked for me…

Getting the right Shared Coordinates in Revit

  1. Start a new, blank Revit model
  2. Link the DXF Centre-to-Centre (this is best way to deal with huge coordinates)
  3. Acquire Coordinates from it
  4. Save your Revit file. You now have the right World coordinates, and a project grid relationship.

Importing the Point Cloud by Shared Coordinates 

  1. Open Recap and import the data. For the ASC data above, on the import settings I used ‘Advanced’, and chose the text columns XYZRGB. I also set the coordinate system.
  2. Export to PCG. Sounds weird, I know. But PCG is a nice reliable container that supports colours.
  3. In Revit, Link Point Cloud, by Shared Coordinates, and choose the unIndexed raw PCG:
     
  4. Revit will now open another dialog, and you can index the PCG file (again) to an RCP+RCS
  5. Link this RCP file by Shared Coordinates
  6. It should be in the right location and related to the DXF coordinate system.

Import any 3D geometry (like Google Earth terrain) to Revit environment via Memento and Recap

Luke Johnson

I got a virtual tap on the shoulder from Brian Nickel over at The Revit Saver today. He said he had been “searching for an answer like this for a year or two, and the guy posted it two days ago overseas! Talk about a neat workflow!“

What are we talking about?

Basically, it is an inventive way to use the mesh tools in Memento, which can consume images, make meshes, and export them to Recap. From there, we can easily get that into Revit for quick site context investigations. The actual implications of this workflow are pretty big… hypothetically, it allows you to use any good series of images to automatically make some 3d geometry and bring it into the Revit environment for further modelling.

Funnily enough, this workflow is just a logical extension of the stated aims of Memento, namely images to meshes as a ‘great companion to ReCap’:

Here’s how arnebjelland describes the workflow from Google Maps to Revit (via Photoshop, Memento and Revit), In short:

  • open google maps in full screen and find your building site
  • rotate the view around your site meanwhile you press «print screen» as often as you wish
  • with dropbox installed, the «print screens» should be perfectly downloaded automatically
  • you should crop your photos to get rid of unwanted text/labels. You can do this by making an «action» in Photoshop and then choose «File» – «Scripts» – «Image Processor» to crop every photos in a folder
  • then you make a mesh in Memento with your cropped photos and export it in RCP
  • import it into Revit and enjoy

Screencast (Brian Nickel):

Images (Brian Nickel):

Steps from arnebjelland at:
Combining Google Maps, photogrammetry and pointcloud into Revit | Digital story of an architect

Heads-up from Brian Nickel at:
The Revit Saver: Combining Google Maps, Photogrammetry and Pointcloud Data into Revit 2016

Using Dynamo and Revit to Convert Point Clouds to Actual Geometry for Use in Tekla or Other Software

Luke Johnson

We often work with consultants using Tekla Structures, but we only recently came across the issue of trying to import point clouds into Tekla. The officially recommended workflow is:
Leica / Trimble scanner — Trimble Real Works — Landxml — Tekla Structures

But what if we don’t have Real Works? Basically, we want to create geometry from point cloud (which we can make into a massive list of XYZ values in a text file). Sounds like an easy job for Dynamo… and in a way, it is.

Here is what worked for me:

  1. Import the source point cloud to Recap, decimate to 100mm grids, and remove all values except X,Y,Z (screencast below). This took me from about 19 million points down to about half a million
  2. Export from Recap to PTS format.
  3. Remove first line in the PTS file using Notepad++ (if necessary). The output should look something like this:

    Here is the Screencast:

    Note: steps 1 to 3 should essentially create a 3 field space delimited XYZ text file with no Intensity, RGB or Normals (sometimes called NEZ by survey people)

  4. Load some family called PointCloud.rfa with a Type called Point (can be adaptive or not)
  5. Use a Dynamo definition to place a given family at each location.
    One of my main concerns was scalability.. How many points / instances can Dynamo and Revit handle here? Initially, I used a method where the definition itself threw away a lot of points in a totally arbitrary manner, using a series of DropItemAtIndex nodes. This got me from about 500000 to about 120000 points, and this worked ok. I ended up modifying the node to allow for a number of ‘drops’ (from 0 to 4). Each drop throws away every second point… Finally, as I was getting all the points anyway, I thought it would be nice to have a Topography creation option. The published package can either create families at each point, make a topography, or both.
  6. Once you have generated the geometry you want from the point cloud, then Export to DWG or DXF
  7. Transmit to consultant

Here is a little readme:
When you first load the package, you should set up the entry data types as per below:

If your text file is space-delimited (as mine was), make sure the delimiter string field actual has a Space in it.

Also, set the two booleans to False (meaning that no families or topos will be created) for the first Run, and set the drops to 4. The “Number of points” output node will give you an idea of how many points are in play at that particular drop level, like this:

If you are running on ‘normal’ system hardware, you probably should keep it to around 50000 geometry creation points if possible. On my Surface Pro 3, it could work with the 30000 points no worries, and my workstation could handle 120000 ok. So, once you have a reasonable number in that output box, you can set the go and place instances and / or make a Topo options to True. I think Revit may struggle with huge points on a Topo, but I was able to place the family instances (with a small crosshair or 3D sphere at the origin) and then export to DWG.

Keep in mind this is a very arbitrary and lossy method – point clouds were never really meant to be wrestled into geometry like this. However, it may help you in certain situations. The Dynamo node has been published but it is very beta at the moment, so of course the usual disclaimer applies: “use at your own risk”.

Package is called Place Family Instances or Make Topography by Point Cloud.

Sample point family for download
 
A note on coordinates and rounding:
This tool currently uses project coordinates. A future revision may offer shared coordinate translation. In the meantime, you could use some reference geometry at project base point and run this tool in an new empty, linked file, then move it into place in project. Related discussion:

Also, it appears that rounding is occurring to 3 decimal places, which is not ideal. Again, this may be fixed in future.

Endnote:
I tried lots of other methods, including POINTSIN and IMPORTXYZ lisp routines in AutoCAD, but oftentimes the dataset was too big, or the input data was not what the routine was expecting.
Some other methods I attempted are below, but they weren’t too successful…
Also tried:

  1. Import points to Civil3D
  2. Convert Civil COGO points to vanilla AutoCAD blocks
  3. Use blocks to generate geometry

Using Civil3D to Convert Points to LandXML for Import to Tekla Structures
Another possibility:

  1. Points into Civil3D (as Drawing Objects in a Surface)
  2. Export Surface to LandXML

Shared, cross-platform, online point cloud viewing and tagging for FM with SCENE WebShare

Luke Johnson

Check it out, pretty amazing stuff. Take a virtual in-browser tour of Sydney Cricket Ground:
SCENE Webshare Cloud

The implications of this technology for facilities management are pretty significant. In some cases, projects may skip the BIM process and go straight from laser scan to FM (where the brief allows for it).

Learn more:

via
Full Webinar for FARO Scene 5.3 and Webshare Cloud | FARO UK Blog

Decreasing the pain with Point Clouds

Luke Johnson

A few tips from Adam Sheather:

  • Change your (Recap) temporary folder locations for conversion 
  • If you need to move cloud file between offices, convert them first to lower transfer times.
  • Ask for ptg files for readability
  • Using Remote Desktop services may be an issue if your card doesn’t support at least OpenGL 3.1 for remote services.

From:
Stuff and BIMs: Point Cloud beginer tips

You may also be interested in my detailed post at:
What Revit Wants: Creating Revit Geometry from Point Clouds (Points to BIM workflows)

Project Tango and the Implications of Commodifying 3D spatial scanning for AEC

Luke Johnson

I was speaking to a surveyor recently and he indicated that he believes in around 2 years time, point clouds will be a “commodity” that almost anyone can produce. Devices to scan 3D spaces are becoming more powerful and more accessible, and essentially anyone can use them.  For example, have you seen Google’s Project Tango?

Along with Recap 360 and other iOS apps for scanning rooms and converting to 3D models, the question soon will not be “how do we scan this?” The question will be:

What do we do with all of this scanned, real world spatial data?

Learn more:
ATAP Project Tango – Google