How a scanner sees a construction site

Laser Scanning and Reality Capture During Construction

The latest software technologies allow for a full digital construction workflow – design is completed in 3D with building models; complete “mesh” and immersive as-built worlds can be captured; and the possibility to simultaneously combine both in one system to visualise constructed elements versus the modelled intent. Models can even be sent to automated manufacture – rebar machines etc.

Never before has so much data been able to be captured to completely record the building inside and out – allowing for the accurate recording of practically any elements position.

In this paper, we will look at why you might consider reality capture in your construction workflow. Additionally, we will look at the following background issues.

The principle reality capture technologies are:

  • Laser scanning
  • Structured light scanners
  • Photography (applied with photogrammetric techniques)
  • Traditional surveying

Sample current software that this technology can be applied to:

  • CAD and building information modelling packages (Autodesk’s Revit + Autocad; Archicad; Bentley et al)
  • Visualisation tools for proposals and animations (Autodesk Infraworks or 3DSMax say)
  • Clash detection tools (Autodesk Navisworks and others)
  • Meshing; surface and volume analysis tools (Sequoia; CloudCompare; PointFuse; Memento; even gaming engines like Unity)

Storage and delivery of data:

  • Capture techniques generate a large volume of data that needs to be archived – consider cloud storage providers (Azure; Amazon)
  • The same data can also be shared and moved – communication networks or couriering of data can be options – there are also streaming solutions to consider (eg Leica JetStream) which allows small plug-ins installed in the likes of Revit to pull data on demand across standard networks in small chunks.

So why reality capture?

Sample output from a laser scanner .. THIS IS NOT A PHOTO .. but millions of survey accurate, coloured points in 3D space

Sample output from a laser scanner .. THIS IS NOT A PHOTO .. but millions of survey accurate, coloured points in 3D space

Reality capture technologies are an order of magnitude (or five!) faster at “pick up” than traditional surveying.

This is easily highlighted with a few simple numbers – a surveyor using a total station can typically measure in the order of ~ 700 points for a busy day .. a laser scanner by comparison can measure up to 1,000,000 points PER SECOND.

This also demonstrates why the data handling technologies and techniques (your game plan) becomes immediately necessary – there is simply a lot of data inbound.

With this sort of speed available, incredibly detailed areas – like a construction site – can be quickly documented for later analysis.

Aside from speed, reality capture allows for analysis to be undertaken at virtually any point in the captured area. The density of data collected is staggering.

This analysis can be carried out by different professions who might have different needs.

When a surveyor measures an area with a total station (or GPS), they apply their own interpretation of what a client wants – they might measure a pipe at a bend, or measure a wall at either end, or perhaps pick up one point on a column and a radius. But this does not tell the whole story – does the pipe dip in the middle; is there a bow in the wall; is the column plumb?

Analyse a floor for flatness with scan data - points will mm's apart

Analyse a floor for flatness with scan data – points will mm’s apart

Likewise, a surveyor charged with doing an as-constructed survey of a slab say, might measure a point every 3 metres, in contrast, the scanner will return a point on the slab every few millimetres – small ponds and ridges can be made immediately apparent; think of all sorts of questions, chances are reality capture can help – how good (straight and plumb) is a wall; what sort of deformation is there in the water tank; and laser scanning is especially useful for hard to reach things like steel beams or dangerous areas of a site (especially train lines)!

By having millions of survey accurate points available for later analysis (via laser scanning at least), the various professionals can analyse objects in incredible detail – as opposed to the interpretation made by a surveyor delivered in a CAD file.

Software can automate certain tasks as well for rapid analysis – a classic function is clash detection – if you have laser scan data, you can “clash” proposed models against the existing structure to find where the existing infrastructure might interfere with the proposed works – you might want to run a new major air duct for example in a ceiling space – the new design and be “checked” before it is even built and this analysis can happen years after the scan was done.

"Clash" point cloud data against proposed models to make sure things will fit before you make them

“Clash” point cloud data against proposed models to make sure things will fit before you make them

Reality Capture Technologies

Like every industry, reality capture techniques and technologies are ever evolving.

Similarly, of the technologies already available, each one has particular strengths and weaknesses when comparing performance criteria like range, speed and accuracy.

The common categories of equipment are:

  • Laser scanners
  • Structured light scanners
  • Cameras (typically via multi-camera rigs)
  • Traditional survey equipment (total stations, GPS, even tape measures still have a place)

Laser Scanning Overview

The name, laser scanner, pretty much describes it – a laser is mounted on a rotating and revolving “head”. The laser measures distance to a reflected surface and the head measures horizontal and vertical angle. When all these data points are combined – the laser essentially hit something and bounced back at a point in 3-dimensional space relative to the “head” .. the head then rotates a fraction to record the next point .. and the next .. until a “point cloud” is built up. This can happen a million(s) times per second.

The whole unit is then moved – lasers can only see line-of-sight – so to see around corners as it were, the unit needs to be moved to fill in the gaps. Laser scanners can be mounted on tripods (terrestrial), aircraft, cars, hand-held (GeoSLAM) – even backpacks.

The location of the scanner positions can be tracked using targets, GPS etc to combine all the scan data into a seamless coordinate system – map grids, site coordinates, AHD are all options – allowing the scan data to be combined with other data about the site like aerial photography, CAD, models and more.

There are various types of scanners that offer different properties – without getting too technical, some are faster (points per second – eg phase-based lasers), different lasers can reach further (not eye-safe), some are more accurate, some produce less “noise” (time-of-flight lasers – a more stable measurement), some are small (less accurate), some are bulky (more accurate but harder to lug around) .. some have cameras built in. Each job can warrant a different or even multiple choices!

Laser Scanning Summary

Strengths    

Accuracy     Terrestrial survey-grade scanners can achieve accuracies in the order of millimetres when deployed with proper survey techniques (eg control network design or when deployed in combination with total stations).

Range     Depends on the unit, but ranges up to 1.5km are possible – note that range does impact on accuracy (the angular component) (100m+ is typical).

Density    Points are record every few millimetres apart.

Weaknesses

Reflections    Reflective surfaces can do crazy things with lasers. There are mitigation strategies that can be adopted depending on the task at hand (like chalk spray, dealing with bulk reflections with scan clean up).

Speed”    This is a relative measure when considering other reality capture techniques – it does take time to capture an area using terrestrial scanning (incl. photos) in great detail. In contrast, photography might offer a faster pick-up albeit with some accuracy caveats. This “speed” measure also incorporates controlling coordinates of the scanner’s position with targeting etc in the field.

Cost    The scanners start ~ $30,000 and can quickly get to $500,000 (say).

Structured Light Sensors Overview

A DotProduct scanner at a refinery

A DotProduct scanner at a refinery

You probably have one of these in your home and don’t even realise – the Xbox Kinnect Sensor is a “scanner”!

“Banded” sources of infrared light come from multiple sources to create an interfering light pattern. The sensors/cameras can see this and determine angles and distances. By moving the scanning head, the software can “track” the changing patterns of light over the surfaces to maintain a coordinate system.

These are pretty limited though – especially range – they only work over a few metres from the surfaces being measured – but they are great in tight spaces like manholes/chambers or plant rooms.

Google have been looking at adding this to phones via Project Tango. Consider doing as-cons with nothing more than your mobile phone – this gives you an idea of where all this is heading!

Structured Light Sensors Summary

Strengths    

Mobile     These units are handheld and easy to use.

Cost    Under $10,000 gets you in the door easily.

Weaknesses

Range    Only works over a handful of metres – you build up a scanned area by capturing all surfaces contiguously. “Closing” off against previously scanned surfaces allows the software to proportion measurement errors. Targets can be applied to the area for control.

Accuracy    Again a relative measure – compared to a laser scanner, this system is simply not as accurate – but when applied for the right type of “subject” (say a pipe ready to be covered up) it is more than likely accurate enough! It comes down to operator experience and user expectations.

Photogrammetry Overview

This “technique” captures an area with photos .. lots of them. The photos need to overlap significantly – in the order of 80% of each photo needs to overlap with others.

The overlapping photos are then “stitched” together to create point clouds and meshes based on a technique called “ray tracing” – this is all done with specialist software – it does this by matching pixels in corresponding pairs to calculate the relative camera positions and then from the camera positions, determine the locations of pixels in 3D space.

This technique is commonly applied to aerial photography – both full size aircraft and more recently drones. It is great for mines for example – but construction sites work just as well – as does doing things at ground level – think façades.

Photogrammetry Summary

Strengths    

Mobile     Start with a just a GoPro.

Cost    Under $1,000 gets you in the door easily, using Software-as-a-Service processing paid per job (eg Recap Pro).

Speed    You can strap a GoPro to a drone and you are starting to look at how many square kilometres you can map in a day.

Weaknesses

Accuracy    Again a relative measure – compared to a laser scanner, this system is simply not as accurate – you are comparing an engineering grade solution where you are chasing millimetres versus something done well can give you ~ 20 millimetre accuracies. These numbers are actually not dissimilar to the comparison between total station derived data and data gathered through survey-grade GPS – so it can still give you pretty good results!

“Plain-ness”    The software needs differing colours and textures in the photography to be able to generate points. White walls, or very even objects can be troublesome – whereas a laser bounces right off, the photography and software algorithms might not “see” there is something in the area.

Hybrid Systems

We’ll take one with all of the above? Microsoft (yes, them – they also do aerial photography cameras for what it is worth) and Leica both do systems with a variety of sensors.

Microsoft UltraCam Panther mapping backpack. 26 cameras for a total of 172 megapixels, 68 megapixel stitched 360o video, 300,000 points per second multi-beam laser scanner, GNSS, inertial navigation, even a Kinnect Sensors, solid state hard drive array for up to 6 hours of data Probably only a  half-million dollars ..

Microsoft UltraCam Panther mapping backpack ..

26 cameras for a total of 172 megapixels, 68 megapixel stitched 360o video, 300,000 points per second multi-beam laser scanner, GNSS, inertial navigation, even a Kinnect Sensors, solid state hard drive array for up to 6 hours of data and probably only a half-million dollars ..

 

Traditional Techniques Overview

Traditional surveying with total stations and GNSS still very much has a place. Total stations rule when it comes to accuracy .. GNSS rule when it comes to convenience so long as you can see the sky.

These techniques can also offer faster pick-up for isolated objects or very sparsely developed environs – it simply may not be worth getting the “bigger” gear out of the car when you have a tape measure in your pocket for a single distance.

It really does come down to the right tool for the job. Having said that though, sometimes you might not know you even have a job or require a particular distance when in the field, and a laser scanner will probably get it saving you a trip back to the site – eg the client wants to know where the overhead power goes and didn’t ask for it the first time.

Reality Processing Technologies

You have your data .. and lots of it .. now what?

Scanning and photogrammetry have been around a long time – scanning has been available since the ‘80’s .. but it is only recently that the software has really started to catch up. So capturing the data is well understood by surveyors and is far from cutting-edge.

Software you might already use in-house can consume a point cloud from a laser scanner.

Combining the data and the tools is less common in the construction industry, but it is rapidly gaining exposure – it is just simply too good once you start getting into it. There is a learning curve – but it is well worth it! The oil and gas guys especially – refineries and oil rigs etc – have been doing this for decades.

Revit and ArchiCAD can ingest point clouds – so you can simultaneously visualise your model data right alongside existing surveyed points. This gives some pretty powerful and immediate returns.

A couple of Workflows

  • Create model objects from point cloud data – you can see the point cloud, cut sections through it, view it by level and so on – and from there you can drag walls out, model window families say, get level information – this process is typically called Scan-to-BIM. There are some additional plug-ins on the market that can help to make modelling even more productive.
  • Use the scans as as-constructed “Documentation” – the scans can be simply put away until you need them again – it might be tomorrow, could be five years. The scanner can pick up anything that reflects a laser. The documentation aspect is especially appealing for things that get covered up – think slab tensioning cables, under-slab pipes, services in ceiling spaces, rebar in columns. Using open file formats (eg e57 means the files will be accessible for decades) (as opposed to proprietary formats).
  • Model validation – you have already gone to the trouble of modelling your new building for construction, so why not make it match what actually ended up getting built? Your model becomes your as-con. This validation can happen in near real-time – scans can be available within hours – allowing you to check starter bars say prior to the formwork going up for the next pour – scanning allows you to develop a visual survey-vs-model quality assurance workflow. Your model can now live on and be leveraged for facility and asset management (see 6D BIM).
Compare as-built versus intent using heat maps against the original model

Compare as-built versus intent using heat maps against the original model

 

Model-vs-As-built validation prior to further construction .. this is a glass roof designed in Tekla - the steel detail design was compared to what was installed. Construction was analysed prior to the shipment of glass leaving China.

Model-vs-As-built validation prior to further construction .. this is a glass roof designed in Tekla – the steel detail design was compared to what was installed. Construction was analysed prior to the shipment of glass leaving China.

There are literally dozens of software packages available for manipulating point cloud and reality capture data. Similar to the capture equipment, all have their strengths. There are also well attended online forums you can leverage with global reach for when you get stuck or need ideas.

The hardest part is letting go of CAD – you don’t need CAD objects to measure, analyse or visualise your infrastructure – you don’t need a piece of paper – you can even do all of this with a tablet in the field, in full 3D.

Just to revisit – the level of detail possible is simply extraordinary – consider the screenshot below from a newly released meshing tool called Sequoia .. how could this possibly be measured traditionally and be described by CAD? This is a 3D model and not a photograph – it was generated from laser scan data.

Scan-to-Mesh

This is a 3-dimensional mesh .. you can measure from this or view it in Revit!

Storage and Networks

The data is comprehensive, as such there is no escape from there being a lot of it required to describe what you see on screen .. and we are talking gigabytes or terabytes for large areas.

Storing the data is likely going to be outsourced to the likes of one of the big cloud providers where you are paying cents-per-gigabyte-per-month to vault this resource.

Moving the data around is another issue – external hard drives are not an uncommon solution to getting the data from A-to-B.

There are however some newer strategies that could be worth exploring. The latest is a Leica product called JetStream:

  • Scan data is acquired, processed and published to a JetStream server
  • The server can be anywhere – but in the same country reduces latency when using the data
  • A “client” connects to the server and essentially starts requesting screen shots – the server processes the request and sends the “points” back for viewing
  • This way – the file is stored once and can be leveraged by multiple “clients” simultaneously – even shared with consultants
  • This can be done over standard Internet connections – even mobile.
  • “Clients” exist for Autocad, Revit, Navisworks, Infraworks etc
  • The point cloud data could be hosted by the surveyor or the building owner – and consumed by the builder, architect, engineer – the scan data could be maintained as a regular, on-going service – even with daily refreshes say depending on the speed of progress.

Conclusions

Hopefully we have provided some insight into the 3D reality capture world.

There is absolutely no doubt, that as sensors come down in price and size, as storage and networking gets faster and as the software tools get more comprehensive, that reality capture will become as pervasive as CAD is now – the evolution path is clear – paper to CAD .. CAD to reality.

We think the time is upon us where real benefits can come from getting adopting some of the procedures and workflows today .. once the concrete is poured – it is hard to measure something!

We would look forward to exploring this with you further in the future.

 

Download the PDF of the full briefing paper on Reality Capture for Construction.

Posted in 3D Cloud Modelling, Building Information Modelling, High Definition Surveying, Point Cloud, Surveying and tagged , , , .

Leave a Reply

Your email address will not be published. Required fields are marked *