Home  /  BIM and ARCHICAD   /  The Making of Point Clouds in ARCHICAD 19

The Making of Point Clouds in ARCHICAD 19

This post is by Ferenc Traser, a Product Designer at GRAPHISOFT. Ferenc has an MArch from Budapest University of Technology and Economics and an MS Arch from DAAP at the University of Cincinnati. Prior to joining GRAPHISOFT, he practiced as an architect, mainly working on historic preservation projects. Having surveyed historic buildings by hand, making point clouds usable directly in ARCHICAD was particularly important to him.

While planning the new features of ARCHICAD 19, we quickly realized that a rapidly developing area in architecture has been missing from previous versions. Known for decades as geographical information system (GIS), this technology has recently made a significant impact on architectural representation and surveying. Remote sensing is now used to analyze not only terrains but also smaller scale objects: streets, buildings, interiors, and engineering details. Architecture firms worldwide believe that instead of spending money on the slow and expensive method of hand surveying, it is cheaper, more reliable and definitely faster to create point cloud models. Now that the hardware has become more affordable for smaller firms as well, it was time for ARCHICAD to gain a foothold in this area.

ARCHICAD 19 signature building as a point cloudThe idea of point clouds is simple: it is an array of points in 3D, each of which is assigned X, Y, Z coordinates and RGB color values. It became apparent to us that a third type of value, intensity, is also used to represent the grayscale spectrum. In some cases, color is simply discarded: the 3D information and intensity values are enough to distinguish between points for modeling. This rather simple set of data is carried by various file types; almost every hardware provider has its own. We chose two that are independent and started to implement them in ARCHICAD.

The point cloud feature in ARCHICAD is a first in many ways. Points as a type of geometry did not exist before; everything was made out of either 2D or 3D sub-elements. Point clouds are enormous in size compared to other objects; millions of points can take up gigabytes of disk space and memory. They are also one of the few multistory GDL elements whose size on the floor plan may vary from story to story. Once you import them, point clouds come in twos, because ARCHICAD generates two files.

One of the formats we ended up using is cleverly named XYZ. Basically it is a text file containing any number of columns (generally six or seven). Since there is no standard, the challenge I faced was to create a method for deciding which columns to use for what values, namely XYZ or RGB. Unfortunately this simple mapping lacks the use of the third set of data, intensity. The other file format, E57, doesn’t need this mapping procedure.

mapping endless rows of numbersDuring the design and development process, we received a test point cloud file from Japan that showed everything in black. As it turned out, all the RGB values were set between 0 and 1. I managed to open the file in Excel and multiplied those numbers by 255. The problem was that in Excel there are only about 1 million rows, therefore opening and tweaking anything larger than that has its limitations. Our small test file had at least twice as much, so when I finally managed to open the point cloud in ARCHICAD, I could see a room almost cut in half diagonally. Only half of the points made it through my conversion efforts.

Another fine example for how different these files can be is that they are not necessarily uniformly scaled. We have encountered a few examples where the Z value of the points was 100 times less than X and Y. This is why we implemented a non-uniform scaling option in the GDL settings.

point clouds are ordinary objects with some extra settingsBefore point clouds, only a handful of GDL elements spanned multiple ARCHICAD stories; except for pipes and trusses, they are all located on a single story. This became significant when we realized that the anchor points of GDL elements are derived from their story-based bounding boxes. Since a point cloud can be in any shape and size, this bounding box constantly changes from story to story. This means, for example, that if you elevate a point cloud, you reshape its apparent size on the floor plan, thus moving its anchor point and its horizontal position as well. After a long debate we decided that a point cloud should only have one anchor point.

When writing the requirements for this project and eventually what ARCHICAD should do with point clouds, we realized that snapping onto points may be out of scope. I am no developer, but I could understand the need for this limitation in the case of millions of points in one single view. I imagined that tracing points with snapping would be the ideal workflow and would improve usability immensely. At the very last moment, it turned out that activating snapping on point clouds is quite simple, and there have been many developments that improved performance in that area as well.

snapping onto points help tracing historical buildingsWhat does the future hold? We will have to support intensity in addition to color. Make performance better and be able to import even larger files. In my opinion, we must enable point clouds for any output: it would be an ideal tool for creating architectural context. Recognizing BIM elements from point clouds may be too far in the future, although companies like ClearEdge3D aim to do just that. But maybe tracing simple surface geometries like planes or tubes can be a good way to start. Whatever direction we take, it all depends on you: how would you like use point clouds in ARCHICAD in the future?

Point cloud images: Bradford College, UK, Bond Bryan Architects, www.bondbryan.com

Did you enjoy this post from someone who helped shape ARCHICAD? Let Ferenc know in the comments. I thought it was super interesting. And I hope it won’t be the last time we have someone from GRAPHISOFT sharing their thoughts about the evolution of ARCHICAD on the blog. Subscribe to the blog so that you don’t miss future posts about the awesome future of the built environment: Shoegnome on FacebookTwitter, and the RSS feed.

Comments

  • January 19, 2016
    reply

    Ferenc:

    Great article. Thank you. We are in the throes of completing as-built record drawings within ARCHICAD for the US Library of Congress using Point Cloud Data (e57 and xyz) from 300 laser scans of the Arizona Fairgrounds grandstand in Phoenix, AZ. Over 100 Billion data points AFTER Faro Scene had cleaned up millions of “stray” points resulting from dust in the air and other artifacts.

    The Faro Scene software is fine BUT it doesn’t really “understand” the process of having to create plan, sections and elevation views and thus its clipping boxes, which enable you to segment very large point cloud environments into more manageable 10-25GB blocks.

    Ultimately some kind of intelligent or interactive element-builder would streamline the whole workflow. However, for existing structures, many of which have very complex structures, uneven surfaces, uneven building settlement, and non-orthogonal walls (more like hand sketches in some cases, being able to capture in plans, sections and elevations, the precise essence of “what is” at a signifiant level of detail is important for entities like the Library of Congress.

    Happy to get you feedback on our AZ Grandstand findings and hope that we can contribute to further development and enhancement of ARCHICAD’s point cloud capabilities.

  • January 19, 2016
    reply

    Before Graphisoft introduced point clouds as a feature I was looking into other 3rd party options.
    Standard PC’s will simply struggle to handle the vast data set of a point cloud, what is really needed is something that simplifies the data set and convert it into something more digestible by a 3D modelling platform. One of those tools as you mentioned was Clearedge3D, I had also been looking at Pointfuse and had introduced Graphisoft UK to the developers to help Graphisoft produce a better tool. Graphisoft at that time were developing their own point cloud tool which simply emulates what Autodesk had already done. However speak to the Revit users also using point clouds, they are suffering the same system capability issues in handling the huge data sets.

      • January 20, 2016
        reply

        One of the problems with FARO Scene (LT and Full versions) is that the whole point cloud appears to be rendered every time you change the viewpoint. SO for relatively small scans with a few million data points it will work quite smoothly but for much bigger scanning jobs the CPU loading and loading on the graphics card can be huge.

        Maybe there could be some ‘take-aways’ for FARO from the 3D gaming environments such as War Thunder and others. These gaming environments are able to render insanely accurate full-blown 3D flight models of aircraft, damage models and super-detailed terrain and water/lake/ocean models on the fly and with incredible smoothness and accuracy – even down to small details such as wind blowing through trees and the sound it makes. If procedural coding could be applied to point cloud files – especially the very large ones that get to between 30 and 200 GB at a time, it might be possible to get to the viewpoint you need very quickly without chewing up all of your CPU space by rendering billions of xyz points you cannot in fact ever see from that viewpoint.

        Just some thoughts…

  • January 20, 2016
    reply

    Indeed the huge data sets present a very big challenge especially for “slicing up” the point clouds into management segments from which useful elements, plans, sections and elevations can be produced.

  • January 22, 2016
    reply

    For the vast majority of applications in manufacturing, architecture, engineering and construction; the place in the data food-chain where we can get most value from our 3D data is when we are able to integrate, manage and manipulate it when in a vector format. Regardless of how our 3D data has been served or shared with us, the sooner that we can convert a point cloud into an intelligent vector data model, the sooner we will reap value from our investment in 3D imaging.

    The 3D imaging industry has been striving for over 20 years to produce automated algorithms that produce the absolute ‘best’ results, in absolutely all cases, with as little manual intervention as possible.

    Pointfuse automatically converts point clouds to vector data models compared to manual or semi-automation methods using a statistical best fit. The vector models that Pointfuse (v2.0) produces are comprised of sets of separate structured surfaces that can be easily manipulated. Common meshing techniques using unstructured surfaces cannot be separated or manipulated; it also cannot be easily styled or labeled with attributes.

    Pointfuse generates a structured model, the vector geometries and primitive shapes produced by Pointfuse v2.0 are separated along their breakline or edge which make them more usable for the user in the software of their choosing as common format like IFC, DXF, OBJ are supported.

    Following the release of Pointfuse 2.0 (March) we will be building on the theme of intelligent models, by defining or categorizing the structure surfaces and would be happy to find out more on end user requirements.

  • January 23, 2016
    reply

    Phil…I’m looking forward to your presentation at Eric Bobrow’s Master of ArchiCAD Summit. I believe just before the Super Bowl weekend? One request: Can you start at a very basic and simple level to explain what is going on here?
    Not all of us are in the same ArchiCAD/BIM league as you and those above.

    • January 23, 2016
      reply

      Hi Steve:

      OK yes, I’ll start with the basics – the why?, the equipment and technology, the process and the hurdles, the challenges, the opportunities etc.

      As far as being the some kind of upper league, if I am in that group then its because I’ve just been around this technology (BIM especially) since its early days in the UK. So it kind of altered my genes a bit as far as what I thought architects ought to be doing, the fluency they needed to have in computing technology and systems, and the professional and public-standing opportunities BIM has afforded the profession for at least 35 years.

      Thanks for the tip though – much appreciated.

      Phil

  • January 23, 2016
    reply

    Thanks Phil…
    At the risk of sounding like I’ve only been into BIM and ArchiCAD starting in 2012 (true), we might start off with coming up with a better term for this stuff then “point clouds”. How does this rather ethereal terminology contribute to anything? And to me, the descriptor GIS (geographical information systems) is not a whole lot better, but it seems that at least many of us are least familiar with this term. Is what we are after IXYZ? That is, intelligent (RGB + intensity) plain old-fashioned XYZ points?
    Truly only two cents worth…Steve N.

    • January 23, 2016
      reply

      I like where you are going, Steve.

      Point clouds is a bit odd but it does describe what millions or billions of XYZ points create when viewed from outside of the cloud. Its like the average weather cloud. Billions of very small spheres of condensed vapor that en masse create the clouds we can see. Wirth thinking about naming conventions though. Good thought.

      GIS emanated from the early days of BIM. In the UK the BIM system I was involved in developing (Applied Research of Cambridge and the Oxford Regional Health Authority, oddly enough) was used by Britain’s Ordnance Survey for producing all of their super-detailed maps at a wide range of scales. Essentially if your system can create polygons and lines and attach alphanumeric data to each one (zip code, population, sales, crime stats, health stats, geo coordinates – you name it) then instead of creating buildings you can create maps. Edge alignment to take into account altitudes, contours and the curvature of the earth all get added to the algorithms that enable polygons, lines and meshes to be created.

      So, GIS came from BIM.

      Thanks for your 2 cents!!!

      Phil

  • January 24, 2016
    reply

    Thanks for the compliment Phil…I will take it!
    Regarding the technology, we need to always remember the guys in Silicon Valley who developed the integrated circuit, the engines of our 21st century digital world. And, of course, Graphisoft and others who provide the remarkable tools described above that can manage billions of pieces of information.

    Here’s another question: Is it too simplistic to just view a “point cloud” as a 3D bit map?
    Until next time…Steve N.

  • January 24, 2016
    reply

    “We will have to support intensity in addition to color. Make performance better and be able to import even larger files. In my opinion, we must enable point clouds for any output: it would be an ideal tool for creating architectural context.”

    Thanks for the article and the valuable comments. I’ve been waiting for Graphisoft’s entry into ‘Reality Capture’ since I purchased a scanner almost five years ago. I guess that I’ll have to wait a while longer before it is as usable as other BIM platforms for managing point clouds. Now that Archicad and Vectorworks have joined the point cloud club (2015), I’m encouraged that advanced imaging tools will continue to amaze us.
    ~Scott Page
    http://www.scottpagedesign.com/

    My article for AECbytes (2012): http://www.aecbytes.com/viewpoint/2012/issue_66.html

      • January 24, 2016
        reply

        Jared,
        I think not. I have tried Archicad 19 with point clouds and found it lacking, but having great potential (I’ve been an Archicad user since v6.5). Much 3D scanning is done in grayscale (‘intensity’ -also called reflectivity), not color, because provides faster scanning with reduced file sizes. I believe that 4gb is the largest scan data file that can be imported into AC19 Most of my projects are much larger. FARO, Leica, Z+F, and other scanner manufacturers should sit down with Graphisoft developers and work out how to make the most of 3D scan data, so that the AEC industry can best use it for their needs and their clients. No BIM platform is perfect, but I’ve yet to see one that has good dimensioning tools for point clouds, allowing one to accurately snap to planes, centerlines, edges etc. I’d like to see what Archicad could achieve, here. Likewise, we need ‘segmenting’ tools (classify, group/ungroup) to dissect complex point clouds as needed. ~60% of architectural practice involves existing buildings, so it makes sense to develop software tools able to manage them easily. As-built conditions benefit from point cloud visualizations, but need better dimensioning tools that work in a CAD-like manner. It’s getting better every six months.

  • January 25, 2016
    reply

    Thank you for your comments and feedback.

    It is important to emphasize that point clouds in ARCHICAD 19 was a first step in the right direction and hopefully others will follow. As a drafting aid it can be very useful: for example you can display now point clouds in floor plan, section and elevations, you can snap onto them therefore they can be traced, you can even override the floor plan cut plane settings and make smaller horizontal sections.

    The scope of this project was to introduce point clouds in ARCHICAD and I think we fulfilled this mission in almost every aspect except for intensity.

    For this first step we have chosen file formats that are independent, we did not want to commit ourselves to any manufacturers at this time. We could not include features like surface detection or managing density for example, those may take a little longer to develop.

    As Jared mentioned, ARCHICAD is ready for point clouds. Surely there are many ways to go in the future, but I believe we created something extremely useful in this release.

    • January 25, 2016
      reply

      Ferenc et al,
      Please have a look at the Laser Scanning Forum (LSF): http://www.laserscanningforum.com/forum/ and the thread related to the Archicad 19 release: http://www.laserscanningforum.com/forum/viewtopic.php?f=22&t=8637
      The LSF is a rich source of information about terrestrial 3D scanning and modeling -worldwide. Without well developed CAD and BIM platforms, 3D scan data is underutilized -we need you to make the most of the vast amount of data that our 3D scanners collect! Just keep it simple and intuitive to use, so that the learning curve is not too steep. I’ve seen great strides in software development since I began scanning, and expect future releases to only to improve.

  • January 25, 2016
    reply

    One of the biggest problems for any BIM system to use the vast oceans of point clouds that go along with a detailed laser survey is how to “swallow” very large multi-gagbyte xyz files or e57 files. Laser scans are done not to provide visual context for new or existing structures but to provide detailed survey information on what exactly exists. Even with 32GB of high speed ram on board and terabytes of disk space available, the ability for any current computing system to “swallow” (aka import) 50 or 100GB files in one go is really limited.

    One of the steps the developers of point cloud data tools need to focus on – I think – is how to make the point cloud data less RAM intensive. This can be achieved by reducing the accuracy of the original scans from +/- 2mm to some larger figure. However, when you do that you lose the accuracy attainable by laser scanning techniques.

    So as far as development activities and priorities go, I think its up both up to BIM developers AND the purveyors of laser scanning hardware and point cloud manipulation tools to collaborate on a) how the data is most likely to be used for producing record, design and contract documents and b) the power of OS X, Linux or WIndows computing platforms that most architects with big commitments to BIM technology actually have available to them. And I’m not talking about what they can afford, I’m talking about what literally is available to them.

    We’d love to have iMacs or Mac Pros with a terabyte or two of RAM on board to make quick work of 200GB point clouds. But that capacity doesn’t exist just yet. My sense is that BIM developers and Laser Scanning hardware and software developers could be looking outside of the their current niches to the world of on-line gaming for ways of visualizing and manipulating highly complex 3D worlds accurately and on the fly. Procedural programming comes to mind….but that’s another topic altogether.

    Graphisoft have made a great start without a doubt. Based on what I know about the innovative power of their development team, I’m sure we will be seeing some very interesting capabilities evolving from the ARCHICAD 19 launch of Point Cloud importing and manipulation.

    My 2 cents

  • February 21, 2017
    reply

Post a Comment