ArchDoc 2013: dawn of the new photogrammetric age

ArchDoc is a deliverable driven (by which I mean students are required to submit a drawing set) module of the masters degree in Conservation at the Raymond Lemaire International Centre for Conservation (RLICC) at Leuven. It is unique in teaching documentation practice as architectural drawing production. Year on year the module adopts selected survey tools to keep the experience relevant. This year one technique has had a resurgence: photogrammetry.

Kastel Arenberg at Heverlee, Leuven, Belgium is the home of the RLICC. Kite aerial.

The  heritage documentation toolbox draws on metric technologies and one of the key objectives of ArchDoc is to teach tool selection on the basis of either direct experience or expert opinion: knowing what is possible with limited resources and which tools to invest in is a big part of the ArchDoc knowledge transfer. The teaching team comprises practitioners from English Heritage,  Carleton University Ottawa and PWGS (Heritage Conservation Directorate) Canada, ASRO Leuven, School of Topographic Sciences Ghent, Polytechnico Turino, University of Milan and the National Technical University of Athens Greece.

L1337378Leuven is 2 changes of train across the channel for me. The SNCB operates some lovely stock like this type AM 75 set snapped at Brussell Zuid.

On returning from ArchDoc 2013 I have come home with the clear impression the camera is by far and away the most powerful and effective tool in the box. Here I summarise the state of the art by looking at some common misunderstandings about technique.

Myth 1: only one tool is needed
It is very common to assume a single tool will solve all information capture needs.  An example of this ‘one size fits all’ thinking is the oft repeated claim that a laser scan equates to ‘digital preservation’, sadly this is far from the case and a combination of techniques is required to get the maximum from any recording opportunity, a scan is as far from a replica of a monument as a set of postcards is: information is needed on condition, material and historical significance from many sources to get the full picture.

Myth 2: Photogrammetry is difficult. This has been all too true until we got a reliable and affordable digital method which doesn’t rely on a stereo viewer for model interrogation. Cheap, high resolution digital cameras have changed everything. Nobody these days has a camera of less than 10mp in their pocket and 25mp cameras don’t break the bank if you are a professional.

Wild P32The days of costly and cumbersome specialist photogrammetric capture are over! The peerless P32 has gone the way of the Avengers look.

Finding accessible and affordable photogrammetric tools has been a problem since the module began in 2000. It has been taught as a ‘watch and learn’ element rather than a full ‘hands on’ experience with mixed results at best. Classical photogrammetry required a complex procedure of capture and processing using costly equipment. So much so that laser scanners have been sold on the story they replace photogrammetry because they are easier to use. I have never believed this having made good use of classical analogue photogrammetric data in 1980s and 1990s and seen successful 3D data resurrection from stereo pairs on projects as diverse as the post fire reconstruction at Windsor Castle in 1992 and the ongoing surface mapping by the Mars Explorer mission from 2004 to date. Stereo image capture is robust, reliable and archive friendly. To me the simplicity of capture makes image based techniques fast, reliable and cheap- the hard bit has always been data extraction: the classical case  meant investment in a very highly specialized skill set and a lack of control by data users: you had to trust the photogrammetrist to trace the lines you needed.

Ortho quality issueGood resolution orthophoto outputs are still tricky. Left is the native image resolution, right the ortho output from PhotoScan. 

By automating image alignment, correlation and surface generation the new family of photogrammetric tools simplifies the process and puts the user firmly in control: it’s easy!

Workflow‘New’ photogrammetry, aka ‘soft copy’ or ‘SFM’ (structure through motion) software such as  Topcon ImageMaster, AGISoft PhotoScan,  Catch123D,  Cirri and Autodesk Recap has matured to the point where digital photographs can be used to generate 3D mesh, point cloud and texture mapped models by simple procedures which can be followed by non-photogrammerians easily with reliable results. It takes patience as the processing stage is machine hungry but the results, provided simple precautions are taken at capture, are worthwhile.

The subject here is the ‘tas-de-charge’  of a late 14th century vault. The transition of the rib profile into the column shaft is too detailed for a TST trace to work well :

Cap shotThis is a hand held shot taken with an APSC format compact camera, it records texture, damage and color and the cost of capture is minimal when compared to a laser scan of the same area.

Cap Photo coverThe 18 image sequence was built up on a ‘try it and see’ basis. First 4 then 8 then 18 images were loaded and the coverage tested by examination of the point cloud output .

Cap Photo cover2aWith the images aligned building the geometry takes a while (it can be hours) before you can see the result but it’s worth the wait

Cap model psan3 Cap model psan2Setting up the photos took about 15minutes and processing the 18 images to produce this model took around 3 hours in PhotoScan.

A little care in managing the mesh and the TIN model is CAD ready:

Cap model cadcad plan viewOnce the mesh is in CAD profiles can be lifted ready for model building, no need to go through the pain of working with the dots in space from the laser scanner!

Myth 3: Scanning is quick and easy. Laser scanning has been around since the Optech patent in the ’70s. It’s not new. Handling the data overload from scanners has been a big problem from day one. Since the advent of the Cyrax 250 scanner it has been pushed hard as as a ‘solution’ to 3D building survey and a replacement for image based capture. Gradually the software tool-set to mine information from the dots has developed but point clouds remain very unfriendly to the CAD environment. On the face of it is a marvelous idea: you can turn up, whack out zillions of 3D points and run off back to the office happy you have bagged all the data ready to meet all your clients wishes at the click of a mouse. And that’s where it all goes wrong: point clouds are just about the worst kind of information to turn into a drawing.  You will have to spend a fair amount of cash (on top of the cost of the scanner) to get a software suite to register, slice, dice and profile the dots. Scanning works but don’t believe the hype that it is as easy as ‘scan and run’ to get a functional documentation set.  Yes the scanner scores on big projects with big budgets, yes the scanner is easier to use than a TST,  but for most day to day stuff the humble camera will do just as well.

Myth 4: Site drawing is ‘old school‘ and not part of the modern tool set.

Drawing 1A good, properly observed and constructed moulding profile is worth its weight in gold. You could get this at a great expenditure of time and effort from a laser scan or from a dense photocover sweep. Drawing wins here on speed, cost and clarity. This one was built up from  profile gauge and close inspection of the cut stone.

It doesn’t matter if you draw like a donkey with a bog brush: drawing is fundamental to architecture, there are tools that can help like the profile gauge and a straight edge but some facility in getting key features delineated is required: mouldings, door linings and window casings need proper attention and that means using a pencil. There is no substitute for understanding how things fit together through drawing.

Myth 5: 3D CAD modelling is easy.

Cap model CAD  WIREIt’s easy to fall into the trap of thinking that the model can come from anywhere, SketchUp models can be lashed together without any metric constraints in no time at all, getting the metric component takes patience and reliable measurement.

Cap model CAD SOLIDThe line out of the CAD model resolves the projection of the curve on curve surface intersection: not a simple matter to fix without the ‘ leg up’ from the surface generated by photogrammetry and the TST trace of the rib paths. Building 3D CAD models is a tricky mix of knowing what to model and how to bring the components together from measured sources be it rectified photo, TST wire-frame, site drawing, photogrammetic surface or point cloud.

The camera is mightier than the scanner. I knew this would happen one day and it was a joyful moment when I realised that 6 student groups out of 6 produced excellent results by use of photogrammetry and the laser scans were used by screen grab alone- the processing and handling of the registered point clouds was simply beyond the resources of the students and the software they can afford to carry. It’s been a while but photogrammetry is back!


About billboyheritagesurvey

Heritage worker
This entry was posted in Survey Practice, value and society and tagged , , , , . Bookmark the permalink.

2 Responses to ArchDoc 2013: dawn of the new photogrammetric age

  1. Excellent post Bill !! I just came across it. You should let us know about it earlier. Just one comment: It is not “University of Thessaliniki” It is National Technical University of Athens (500km south of Thessaloniki!!!). Andreas

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.