AECbytes Viewpoint #66 (November 29, 2012)
Scott Page, MArch, U.C. Berkeley
Founder, Scott Page Design
If you are unfamiliar with the terms; LiDAR, 3D laser scanning, point clouds, meshes, and NURBS, this article might scare you off, but I would hang on and hopefully learn something about reality capture from 3D laser scanners and the “point clouds” they generate. I will focus primarily on the use of point clouds into BIM platforms for architects and engineers.
LiDAR is the umbrella term for detection technologies (aerial, mobile, terrestrial, etc.) that employ the principle of radar (pulsed electromagnetic waves) but use light, generally from a laser. 3D laser scanning is essentially the swift capture of three-dimensional information reflected from an object or surface to a light sensor. It creates a 3D construct commonly known as a “point cloud” made from multiple scans that have been unified through a process of “registration.”
3D laser scanning is as much about precise measurement as it is about stunning visualization. Terrestrial scanning has similarities to both surveying and photography (it begins with the tripod). The actual scanning is the least complicated part of the process— the most challenging is the smooth export to BIM. Infinite views, from any vantage point, are available from the unified point cloud. Once the 3D point cloud data is consolidated and exported to a CAD or BIM platform, traditional A/E deliverables such as 2D plans, elevations, and sections can be readily extracted, as shown in Figures 1 to 5. While 3D models depict ideal conditions, 3D scans reflect the buildings as they actually are: seldom perfectly straight, level or plumb. 3D modeling is simplified using point cloud data for referencing, but the point cloud itself can serve this purpose, saving many hours of digital model building.
Figure 1. An axonometric view of the scan of 1204 Mason St. San Francisco (FARO Scene 5.0). While the scanning was all done from the sidewalk, infinite views are possible once the point cloud is unified. Trees that obscured the building’s facade, were made more transparent for improved visualization.
Figure 2. Elevation and perspective views of the scan at 170 Valencia St. San Francisco (FARO Scene 5.0). Both images are from the same point cloud; the vantage point, background color, transparency and resolution have been changed. The cars, freeway, fence, and palm tree were deleted.
Figure 3. Laser scan of apartment plan (FARO Scene 5.0). Complex room layouts are precisely captured along with contents and reflected ceiling images.
Figure 4. Section of laser scan of structure in Berkeley, CA (FARO Scene 5.0). Note the floor slope, structural framing, and complex roof geometry.
Figure 5. Section/Elevation and axonometric view of laser scan of Berkeley City Club, Upper Dining Room and Hall (FARO Scene 5.0).
GIS (Geographic Information System), ‘remote sensing’, stereo photogrammetry, and LiDAR are overlapping, competing technologies that very likely will find their place under the mother platform of BIM, working in concert to capture the world as it currently exists. Today’s architectural training is mostly directed towards creation, not the documentation of existing conditions. The “new” is simply more alluring to young, creative minds. It is far easier to create something new in BIM than it is to accurately model the terrain and detailed features of existing structures. Documenting as-built conditions has historically been a tedious, often inaccurate process. 3D laser scanning offers a fast, accurate, albeit costly solution to this problem. The initial expense of the scanning is offset, however, by the speed, quality and quantity of the visual information produced. Scan data produces enormous file sizes. Accordingly, hardware must be up-to-date in order to accommodate the deluge of scan data. A minimum of 8GB of RAM (one cannot have too much RAM), 64 bit multi-core processors, and the latest generation graphic cards are required to handle point clouds efficiently. Fortunately, hardware costs have plummeted in recent years, while the software is racing to keep up with the processing advancements.
My first encounter with 3D scanning arrived in 2008 during a routine site visit to a house under construction for one of YouTube’s co-founders. A scanning service was employed to capture an area of complex geometry for a stone fabricator. Having used 3D modeling software since 1989, the value of laser scanning was immediate: Here was a way to capture as-built conditions into 3D, without having to laboriously transfer and interpret countless measurements and photographic references into CAD. It took several more years before I looked into 3D scanning, as applicable in architectural practice and BIM.
With the purchase 3D scanning equipment in 2011, I began testing scan data in a variety of software platforms including Revit 2013 (Figures 6 and 11), ArchiCAD 15 (Figures 7 and 8), Sketchup Pro (Figure.9), PointCab (Figures 7 and 9), and Pointools. Rapidform XOR, primarily a CAM application built around point cloud technology, is also a potential platform for architectural projects due to its robust tool set for meshes and NURBS. Transforming point clouds into polylines, solid models, and surfaces requires new training, of course, but it is both doable and rewarding once the work-flow is established.
Figure 6. The laser scan of Pearson Sound Theater, Berkeley, CA, imported into Revit, which now has the powerful ability to directly import unified 3D point clouds, but at the steep price of over-decimation. (Compare to Figure 7 below).
Figure 7. The laser scan of Pearson Sound Theater, Berkeley, CA in ArchiCAD, imported as Orthophoto (TIFF, JPEG, PNG) imported via PointCab for tracing and dimensioning. (Compare this to the Revit image of the same structure shown above.)
Figure 8. 2D orthophoto (TIFF, JPEG, PNG) imported into ArchiCAD for polyline tracing and dimensioning.
Figure 9. 2D plan and elevation orthophotos imported into SketchUp via Pointcab for tracing, modeling and dimensioning.
Figure 10. Kubit’s ‘VirtuSurv’ connects to any CAD interface including AutoCAD, AutoCAD LT and IntelliCAD. VirtuSurv links to Windows based programs (Excel, Word and more), for transferring scan data (planar views) directly to a spreadsheet.
Figure 11. Revit has the powerful ability to directly import entire 3D point clouds using its .pcg format. Once imported, the point cloud can be treated like any Revit object.
Autodesk and Bentley have made the greatest inroads toward importing 3D scan data (point clouds) into their A/E product lines. Trimble’s recent acquisition of SketchUp ( see Figure 9), the fastest growing modeling software, and Tekla in 2011, provides a tantalizing window into the realm of possibility where specialized hardware and software are potentially crossed fertilized. (See AECbytes Newsletter #59.) Mergers and acquisitions don’t always result in product development, and may, in fact, stall promising technologies in favor of existing, profitable wares. So, it remains to be seen where point clouds will “make rain.”
3D imaging is a disruptive technology and has yet to find a settled position in the A/E world. With time, the process will be smoothly integrated into BIM platforms as file export standards such as the ASTM E57 format gain general acceptance. At present we must contend with a confusing mix of proprietary, ad hoc, and domain specific file formats. Just as DWG/DXF formats eased interoperability between many CAD programs, the fledgling open source E57 format will hopefully improve the importation and exchange of 3D scan data to most BIM platforms. The move to BIM is increasing the demand for 3D scan data from as-built conditions. In fact, the U.S. General Service Administration (GSA) has recently mandated the use of scanning: “Every federal facility must be documented in 3D.” “GSA's Office of the Chief Architect (OCA) is currently encouraging, documenting, and evaluating 3D laser scanning technologies on a project-by-project need basis. 3D laser scanning has become a prominent vehicle for acquiring building spatial data in three dimensions with high fidelity and low processing time.”
Scanning data and BIM are potentially well matched for as-built documentation—all that is needed is a common language and uniform tool sets to complete the marriage. 2D deliverables derived from point clouds will only improve the quality of traditional plan sets essential for planning departments and builders, even if the additional 3D information is not directly accessed or immediately appreciated.
Presently, we have an abundance of third party software solutions, plug-ins, and competing CAD platforms—a veritable Tower of Babel of technology. It’s nice to have the variety, but it’s a confusing, inefficient, and unsettled mess to work with. Equipment makers have added limited drafting and modeling tools to their registration software, but most choose to cooperate with the major A/E software providers for drafting and modeling solutions. Any program that has 2D capabilities, imports point clouds, and has the tools to limit and slice the point cloud can satisfy the needs of most users. Indeed, ArchiCAD and SketchUp, which currently offer no serious plug-ins for point cloud insertion, have little problem accepting scaled photographs (called “orthophotos”) that can be easily traced and dimensioned for 2D deliverables (see Figures 8 and 9). This solution, while lacking the flexibility, elegance, and power of direct point cloud insertion into CAD/BIM platforms (such as AutoCAD, Revit, and Microstation), does the job fairly well.
Object recognition/extraction (pipes, walls, slabs, roof planes, terrain and vegetation) from point clouds is the “holy grail” for point cloud software developers (For example, see ClearEdge3D). With object recognition, converted point clouds elements would be treated like other vector based objects, potentially removing the need to model everything to begin with. The modeling process would essentially be partially automated. Within five years time, I expect to see ‘intelligent’ dimensioning tools that recognize planes, edges, centerlines, and corners within point clouds, in a manner similar to the tools found in the best BIM applications. If point clouds are all about millions of precise x, y, & z point positions, then let’s have user-friendly, intuitive dimensioning tools. Let’s start the project with actual GIS positions (if available), orient the scans to them, and finally build the 3D model from this foundation—not the other way around.
I would also like to see every point cloud have the use of a “limit box” with the power of a typical CT/MRI viewer: the ability to automatically cut up a building into multiple vertical and horizontal slices (tomography). Point cloud derived building sections already have the capability to accurately capture geometry, structure, and deformation, but I would like to see this feature leveraged for animation purposes, within the BIM platform itself (for example, see this YouTube video).
Terrain modeling, always a difficult task in Revit and ArchiCAD, would greatly benefit from 3D scan data. Landscape features (such as trees, plants, fences, hardscape, and slope changes) are routinely captured along with the building scans. All that is needed are the proper mesh tools to accurately contour the 3D scan points. It is even possible to digitally separate existing vegetation from the land below it, as shown here.
Lastly, I’d like to see the 360º panoramic photographic views captured from the laser scanning positions linked to the BIM model in some manner. Generally, the panoramic and planar views (jpegs) remain with the point cloud registration software, and are completely overlooked once the point cloud has been exported to BIM. Rich, photographic information will help the modeling process, while reducing the need for (costly) return site visits for verification purposes. Kubit’s VirtuSurv (shown in Figure 10) makes use of this visual information (planar views) for direct take-offs into Excel spreadsheets (as shown in this video). There is more to 3D scan data than the x, y, &z points alone.
Revit (see Figure 11) looks to be the most promising platform for point cloud importation, but is currently a disappointment from its over-decimation of the scan points (shown in Figure 6), its tepid object recognition capabilities, and its lack of point cloud editing tools (to select and deselect subsets of points). With Autodesk’s recent acquisition of Alice Labs(in 2011), there is the hope for robust point cloud editing tools within Revit, and across the entire Autodesk product line. It remains to be seen which features Autodesk will adopt and maintain over time. Likewise, Bentley’s purchase of Pointools (also in 2011) adds great potential value to its line of A/E software products. Once BIM software developers perfect point cloud object extraction, data compression, and dimensioning, we’ll have an extremely powerful platform that melds the actual world with the fertile dreams of creative designers.
Scott Page, M.Arch 1990, began his design career just south of the Norwegian Arctic Circle, preparing countless 4 meter bore holes for blasting in solid rock. Intrigued by the limitless possibilities of design, he completed his formal education at UC Berkeley, and switched his medium from stone to wood. His career evolved from furniture design/fabrication to architecture over three decades. His recent interest in 3D imaging was a natural off-shoot from extensive architectural modeling. He currently resides in Berkeley, California.
Scott can be reached though his website:
Note: The views expressed in Viewpoint articles are those of the individual authors and do not necessarily reflect those of AECbytes. Also, AECbytes content should not be reproduced on any other website, blog, print publication, or newsletter without permission.
Have comments or feedback on this article?
Visit its AECbytes
blog posting to share them with other
readers or see what others have to say.