This week I began my pilot case study, beginning to reconstruct ‘House 7’ at Skara Brae with a view to testing and refining my methodology before I begin work on the whole village…a process which inevitably turned out to be easier said than done!
The house itself is off-limits to the public and is protected from the elements by a covered roof, though the structure can be explored in some sense through the replica house next to the visitor’s centre. My methodology uses the laser scan data as a basis to build the reconstruction, with the aim of remaining as ‘objective’ and accurate as possible structurally speaking. In addition to this, having the original scan mesh on a hidden base layer of the model will theoretically mean that the site as it stands today can be digitally preserved within the file and can always be returned to. With the more subjective elements of reconstruction being mapped throughout the process of modelling on additional layers. Ultimately providing a kind of ‘digital narrative’ of the interpretive reconstruction process as it develops over time – decisions made, conclusions reached and so forth.
But…of course before I can even begin to think about mapping the interpretive process, I have to work out how to model the scan data. My initial instinct was to use the scan data as a basis to create solid geometry, and build the house brick by brick with a handful of generic shapes pulled into place. Although this is arguably the best method in terms of creating a photorealistic model with accurate lighting and materials, it is not especially true to the scan data.
The issue I have with using the scan data directly is firstly its complexity for photo realistic rendering due to the extremely high polygon count, and secondly that the data itself is riddled with holes. Holes in scan data always occur for a number of reasons, in this case it was due to the relatively organic nature of the structure being scanned. Structures within the house cast shadows , collect dust (which can reflect the laser) and are riddled with nooks and crannies which the scanner beam can’t always reach.
Capping and bridging gaps in the data is time-consuming, but is the method utilised by the Scottish Ten team when visualising the scanned sites. However, the worry I have with this approach is that I will be reconstructing elements of the house as it was during Neolithic occupation, so I will be merging mesh and solid geometry – which could get very messy and be very problematic when I come to light the scene.
With these issues in mind I set about experimenting with different methods of modelling the scan data. The top two images are the scan data itself with holes capped (the image on the left has also had a ‘smoothing’ modifier applied to its surface). Although this method remains true to the scan data, there are numerous rogue polygons poking out of the surface causing light to react strangely, and there is no real sense of depth in the gaps between the bricks. In addition to these issues, the polygon count for this tiny section of wall alone was huge! So not a very sustainable method for my purposes.
Following a discussion with my old Southampton MSc buddy Grant Cox I tried modelling the walls using a technique most often used by gamers to make low poly characters appear more detailed. I created a low poly ‘wall’ and used the high poly scan data as a cage for a normals map (bottom left image). I’m relatively happy with the results, as it was very quick to create an impression of the wall, although there is still no sense of space between the bricks.
The final method I experimented with was suggested during a brainstorming session with a couple of the guys from the Scottish Ten Team (Al, Craig and Jared). I used the freeform modelling tools in 3ds max to ‘step build’ a simpler mesh snapping to the faces of each brick and a ‘shell modifier’ to give them depth. I need to perfect my technique a little more, but this is looking more and more promising!
This is never going to be an easy task, but it feels like I’m making progress…I think! Modelling scan data in the way I need to is never going to be easy as in some sense I’m approaching the task backwards, beginning with complex data rather than starting off simple. I’m starting to wonder whether my methodology of digital preservation together with interpretive reconstruction should be two separate things, although issues will always arise when objectivity and subjectivity collide.
I’m off on a data collection trip to the Skara Brae tomorrow, so no doubt I’ll have changed my mind again about how to approach these issues when I return from the islands next week, such is the exciting nature of research!