thoughts, ponderings, experiments & recent things
A simple and obvious trick for catching those fiddly fine details as part of your photogrammetric modelling.
3D models, photogrammetry. All the rage these days. Sometimes though it’s just not possible to model those very fine details you want to record.
There is always the time & data volume tradeoff – it’s not practical to model everything at the microscopic level, or the material is ill suited for photogrammetry, or maybe the available equipment isn’t quite up to standard, or perhaps you just stuff up a shoot.
Whatever the cause some details that are easier (and better) recorded with techniques raking light, hand illustration or RTI.
So why not do both?
You could of course try and model an object entirely in raking light, and again in diffuse light, but that wouldn’t work very well as you loose information in the shadows and highlights that you want from raking light. But it’s quite simple to take a few extra shots during your modelling that you can use as textures later on.
A tripod (or sturdy setup of chairs and bags of lentils) to hold the camera is essential, but the principal is the same regardless of whether you are using a turntable setup or not.
Every so often in your imaging sequence, take a shot in your main modelling diffuse light, then stop, making sure the camera and object do not move. Then turn off your lamps (or change camera settings) and use a 3rd light (or better external flash) and take your raking light photographs. You could even do a full RTI shoot if your setup leaves you enough space to work with.
Then carry on as you were.
When modelling, exclude these images from the process until you get to the texturing stage. Rename these files to the name of the diffuse images and turn off all other images then build the texture. With care you can texture the whole object in this way, and even trace off fine details and save new images to map onto the model. You may find UV unwrapping the model and compositing textures externally helpful to get best results.
You can even trace features and save new textures that can be included in the same way. What’s nice is it creates a very simple record you can revisit that sits well as part of your raw photogrammetry photo archive. And of course you can export each UV mapped texture for your final model for archiving and sharing too.
When preparing for field work, I can whole heartedly recommend not doing this to your macro lens…
some time, quite some time later…
cleaned, re-greased and reassembled.
Nikon 105mm macros lens from ~ 1983.
cleaned with 50:50 ethanol water
greased with TF2 lithium bike grease
take notes and photos
don’t loose screws
More detailed. Sketches to come later.
remove back plate (note spring when replacing)
remove rubber grip
note position of the 3 little screws that are in narrow slots. remove
set f-stop to 2.8 and un-hook/tilt f-stop ring (lift one side – only one will lift)
twist lens round and round and round till it unscrews – try and note the relative positions off all segments at this point.
twist the nose part till this comes out of the big spirally by that was under the grip.
clean lens elements and blower liberrally.
clean spirally bit (helicoid) with different cloth.
paint on grease with artists paint brush
screw the nose back in (starting from the same position as it was just before it came apart)
screw nose into spiral till thread is invisible but before the square bump goes out of view
pier through the back lens to align the metal guide rails (you’ll see them in the back section) line up with hold areas in front.
make sure f-ring is tilted again
twist central spiral section so that it screws into the back, and the nose moves without turning.
slot f-stop ring down and check it moves the aperture
screw it all down to most compact state
put the 3 small screws back where they were and double check your focus
replaces all other bits.
to clean between other lens elements, such as front you need 2 flat head precision screwdrivers and a cross handle (or vice). Well worth it if they are looking foggy.
More ZooMS coming when I’m back with the main hard disk and requisite graphics. For now. Note to the wise.
If you’re planning some Near Infra-Red Reflectance Transformation Imagine (NIR RTI), check your balls before you start, not all shiny black spheres are alike. It’s also worth to prepare some background material – the image above/below is actually black velvet.
The parchment sample, safe and snuggly secure wrapped in eraser particles can be kept without risk of decay for some time, but to get the results it has to be extracted from its polymer home. To do this a solution with the enzyme trypsin is added and the sample warmed to the enzymes functioning temperature ~ 37°C for 4 hours. If 37°C sounds a familiar temperature it’s no coincidence as trypsin is a common enzyme used in digestion.
The animation may look quite straightforward, but for me it is probably the most bamboozling stage to think about… so there’s a lot of writing below.
There are a couple of things going on here. Firstly the collagen strands are being released from the eraser particles, and secondly these strands are broken apart, first into gelatine, and then cut up by trypsin into parts called peptides, but it’s a little more complicated than that.
As we learnt in the previous post the collagen is left clinging to the eraser as a result of static electric charge. When the solution is added to the sample this attraction and bonding begins to give way. Adding the liquid surrounds the samples in a dense sea of molecules, bringing with them their own charges and electrons that interact with the proteins and polymer, effectively neutralising the charge imbalance between them. Heating the sample also has an effect. Introducing energy into the system causes the molecules to vibrate, jig around and knock into each other (try this if you like) causing the eraser to relax, denature and uncurl and helping the collagen and polymer to separate and diffuse into the solution to be digested.
The fibres that make up skin (such as parchment) are thick bundles of collagen proteins which are themselves tight bunches of long chains of amino acids and peptides (short amino acid chains) that are bound to other long chains in a triple helix form. To get a more visual description of this last sentence at look at this video which was a helpful aid to the sequence used in the ZooMS video, or the 3D model on this page.
The warm solution acts as a solvent to the collagen, breaking the hydrogen bonds that hold the triple helix together and splitting them into long gelatine threads – the clear gooey stuff used for making gummy bears, or the jelly in pork pies. At the same time that this is going on the enzymes, as enzymes do, bond to particular points along the length of these chains, splitting them further into the peptides that are needed for the ZooMS method.
Different materials will respond to the heat differently, the molecules in the liquid are relatively free and able to move around and bounce about while those in the protein chains are tied together by attractive forces, with the long chain only able to jiggle around like an angry worm. Eventually some of the bonds will break as attractions between molecules are interrupted by things in the liquid (such as enzymes). To abstract this a little, imagine the scene a bit like 3 giant parallel conga lines that get caught up in a rave, at first they will stay together, but before too long the more freely dancing crowd will lead to lines parting and eventually breaking up into dancing couples.
At the end of this process we’re left with few bits of eraser sat in a soup of peptides and enzymes that can be run through a very fine mesh to sieve out the larger molecules leaving us with only the peptides.
There’s really quite a lot going on here and at sizes that are tricky to get your head around. To stuff it all into a short animation is perhaps a little optimistic, so like the earlier sequences there have to be simplifications.
Lets start with the scales. Now, some molecules can form very large structures, but even so it is very difficult to see their form and requires much more magnification than the kind of SEM that can show the collagen strands in skin. While scanning tunnelling microscopy (STM) can let us ‘see‘ (and animate) individual atoms and thus structures of some molecules, most of the images we have of molecules are models, theoretical constructs developed to represent the behaviour of molecules. As you can see from some of the images and videos linked above these models are often simplified to different degrees depending on the interaction that is being explored or shown. This is true here, the triple helix is portrayed as 3 simple long spirally tubes at the start of the sequence, but the closeup window shows a series of interlinked angled objects (the peptides). In reality these are all shapes of convenience. The lines between corners of the angular shapes represent invisible forces binding atoms together and these atoms themselves are not hard blobs as often depicted, but clusters of even tinier particles orbited by a system of electrons… kind of… if you’re interested Richard Feynman is a good in to the bamboozling world of the frightfully small, paragraph 4 of this page sums it up quite well.
The shape of the triple helix too is simplified. For a start the video only shows the disentangling of a single collagen triple helix separating off the eraser, when in reality there are great bunches of these proteins unravelling. To give an idea of the kinds of scales, imagine focusing on a single tiny strand of hemp in large ships hawser as it breaks and frays in the midst of a heavy storm. The difference in scale between the unravelling fibres of that single thread, the thickness of the rope and the ship it is breaking free from are immense, not only that but simulations suggest the collagen doesn’t simply unravel but come apart and rebind repeatedly before finally coming undone. All in all a very chaotic and confusing scene.
There is also no depiction of enzymes in the animation, instead the focus is on the breaking of the gelatine into peptides, introducing a shape that is returned to in the next animation. This cutting action is important and enzymes are often depicted as scissors chopping up molecules despite looking quite different. If we were to be strict on scale too our view would be obscured by the myriad molecules of the solution everything is floating in.
style and assembly
The desaturated look is continued from the previous sequence, another nod toward the microscopic world where colours begin to loose their familiar meaning. The style is not an accurate depiction but one that tries to simplify things (see above) in a way that is approachable, but retains some of the drama of the event and the key features of the collagen-gelatine-peptide sequence. Computer based molecular modelling that produces more accurate depictions of processes is quite advanced but requires specific software and produces results that are a little harder to understand for the uninitiated. Try an internet search for videos of ‘molecular dynamics‘ to get a better idea of what these molecular simulations look like.
The objects in the scene are very simple, 3 spirals, a blob and a few randomly moving spheres in the background, but behind the scenes are hidden objects that bring it all to life. Central to the animation are the collagen strands. These simple spirals appear to wrap around one another but are kept in separate layers so that they can’t interact. Each spiral has its own rig of forces that pull it apart. Using forces adds unpredictability to the animation where a few subtle changes of settings produce very different results. There are 3 forces for each each with a limited range of influence. The first is closest to the spiral and approaches from below, tugging at the spiral and pulling it out. Too strong and it causes the whole spiral to collapse, working alone and the thread gets in a bunch that follows the force along. This is where the second force comes in, intersecting with this first, it plucks the thread out a little only for it to then be spun out further by a third ‘turbulance’ force. The forces of each spiral are set slightly differently and follow subtly different paths to give an uneven unravelling.
All the while the chain jiggles. This is done using armatures – the same kind of animation bones used to move the sheep. There are 2 sets of armatures, the first a series of short repeating bones that jiggle back and forth and move the spirals, the second a large armature of longer bones that bend this first set, turning it from a curve to a straight line.
One caveat of using this force and armature approach is that the forces don’t like to move around a changing curve, so instead of having them move round in a changing spiral that comes away from the blob of eraser, the forces move in one while the spirals straighten up into their path and the blob moves off. At the same time the camera rolls and turns to give the impression that the spirals are moving as much as the blob.
The last thing was to give a sense of the threads dissolving into the liquid, a simple camera trick of setting a shallow depth of field and adding a touch of motion blur, throwing the wide moving close-to camera sections into a milky blur.
I’ve included an early draft of this unravelling in a video below and a few screen shots with the forces drawn on to explain.
The triboelectric effect discussed in the previous post is shown in another animation – a closeup up of someone gently rubbing the eraser over the surface of parchment. Once again, the emphasis is on how very low impact the sampling technique is, and that it is only tiny microscopic strands of parchment material – ones which are already loose – that are collected. This is in stark contrast to other methods that may require segments of parchment to be cut out and sent away.
what’s going on
Parchment – a form of processed animal hide – is made up of millions of strands of long chains of a protein called collagen all interwoven in a tangled multilayered web. As the eraser builds up a charge and passes over these strands it attracts them toward it. However, most of the chains are woven in place by one another, so while any loose ends may reach up like excitable grass as the eraser approaches, they fall back down again when it passes away. Among this large web of fibres there will be some that are loose. These are drawn up form the surface by the charge and cling to the eraser surface.
The polymer of the eraser is very soft, and while the gentle rubbing does no damage to the parchment, the eraser begins to break apart along the edge resulting in the familiar crumbs of “erdu” that are still covered in the loose collagen strands, making it very easy to collect these microscopic samples.
What’s more, most of these small blobs of eraser are actually tiny little spirals, the curling most likely a combination of the surface charges along with differential stretching of the polymers that make up the eraser. This type of curling may be familiar with if you’ve ever curled the end of a bow on a present by dragging scissors along a ribbon. These tight spirals trap the proteins, locking them away from light, oxygen and other things that may cause the sample to denature over time and so greatly increases its shelf life.
styling… why its all grey.
First of all, lets talk about colour. Although parchment looks yellow, collagen strands themselves are grey. This is the first reason the sequence is in black and white. The second relates to the scale. When looking at structures this small in the real world we need to use techniques such as SEM – Scanning Electron Microscopy. SEM detects surface structures using a beam of electrons, rather than magnifying an image in visible light like optical microscopes. As a result they cannot produce colour images in the conventional sense but result in detailed grey images. Beautiful coloured SEM images are increasingly common these days, but they are coloured in, either for aesthetic effect or as a key to show specific properties.
This animation also has a lot of grain or noise compared to the others. This is a reference to the static that is so important in the collection process. Originally the sequence had more noise added to it in post processing, but was removed due to problems with compression artefacts when preparing the video for the web. A practical upside to this style was to speed up rendering time by lowering the number of samples the blender rendered carried out. Some of the images on this page have been rendered with more samples to give you a now-noise view of the setup.
At this scale the real situation would be much more complex with other substances littering the parchment surface such as dirt, dust and bacteria (but not such that it affects the results of the technique). The eraser too is quite stylised. If we were looking at it at the same scale as the collagen we would not have a sharp edged smooth surface as depicted in the animation but something more pitted and uneven, perhaps like this SEM. The size of the eraser tail would also be much longer and the pieces that come off would be much more complex. But all these things are secondary, the aim of the animation is to show what is happening to the collagen in the parchment being sampled.
the making of.
The collagen strands and surface of the parchment are made from the same type of particle simulator used for the sheep hair. Layers of simulated hair were stacked on top each other to provide depth, and brushed in different directions with one layer allowed to interact with the attractive force field that follows the eraser.
To model the eraser tip, bending and stretching as it drags over the surface was a little more tricky. A softbody cube was created, the mesh adjusted to allow sufficient deformation to get the desired effect without giving the software too many calculations to work out. In the end the cube was subdivided toward the tip and the material weight painted to prevent the top of the model from deforming and reducing the number of faces involved in any calculations. There was still a lot of tweaking required to get the right kinds of tension, spring etc in the material and a few other touches including as a wave deform effect was added to create a subtle vibration across the surface to hint at the friction between the two surfaces as they bond and separate.
A simple trick was used to create the curling spawl that comes off the eraser – a method where you deform one object to the shape of another. In this case a flat cube subdivided to give places it can bend along, smoothed then pushed into an invisible coil shaped box or lattice.
Some images showing the sequence from different viewpoints and the underlying meshes are shown below along with some early concept models. The next instalment looks at unravelling the collagen.
This is probably the most important part of the ZooMS video, and even the ZooMS process. The original concept was by Matthew Collins of the University of York, keen to highlight how truly gentle the sampling method is. It made its first outing to coincide with a publication about the new process, you can see more with these links.
Most people will be familiar with the buildup of small electrostatic charges in their daily lives, such as the crackles and pops you get when taking off a wool jumper, but the most evocative example has to be the balloon-on-hair trick. The creation of the static charge when rubbing two materials together is called the triboelectric effect – tribo coming from the greek word τρίβω to rub. What is happening to cause this build of charge is the two materials, balloon and hair (or eraser and parchment), undergo a temporary bonding when they touch. During this bonding the electrons are shared across the interface. Pulling the materials apart breaks the bonds but not all of the electrons that have been shared out are returned to their original owners, with one material leaving with more electrons than it started with and the other with fewer. In this instance it is the balloon that comes away with an electron profit.
Rubbing back and forth, repeats this bonding and breaking process over and again, which builds up the amount of charge imbalance.
In a way you could think of it a bit like taking 2 sprinkle topped cakes and touching them together, but one of these cakes is stickier than the other. When you pull them apart some of the sprinkles get swapped from one cake to the other.
The sheep in the video is a wensleydale sheep, a particularly hairy bread of Ovis aries picked out by Matthew for its long curly locks of wool that resemble the spirally molecular form of collagen – the protein that is sampled in the ZooMS process. More on this in later posts.
The model of the sheep is very basic. Fundamentally it is a group of simple spheres and cubes that have been lightly sculpted and all tied together with some armatures – or animation bones, not the best way to build a sheep, but effective for this purpose. The hair of the sheep is a mix of different particle simulators. These particles – the hairs – can be affected by forces, wind, gravity, attractive forces etc, making the animation of the hair relatively simple. The only trick is to make sure the balloon only moves the hair you want it to move, and that when it falls back it falls over, rather than through the sheep. This is done using invisible collision objects.
For a more advancedBlender.org sheep, check out Frank, by the fine folk of the Blender community in their 2015 animation short Cosmos Laundromat.
At last, the ZooMS video has come out of the virtual work shed and released into the wild. ZooMS stands for – Zooarchaeology by Mass Spectrometry and is a method for identifying animal species from tiny samples without damaging the original sample. You can find out more by watching the video here or going to these articles.
To accompany the video I’m writing a few posts about the making of the video, specifically the animations, going into more detail about the science behind each animation as well as how they were made and some of the stylistic and artistic decisions taken along the way. The models themselves are all quite simple, but hopefully this paradata will be useful or interesting to some interested, even if the methods used may not be the most efficient or orthodox.
The video is based around sections of footage showing the different stages of the technique, from the very easy sample collection, to the scientific nitty gritty in the lab. Without an on screen presenter it is this lab footage that really sets the overall pace for the video, a video which aims to provide a concise overview of the method to an informed audience of heritage professionals who may consider the using it on their own collections.
At each step of the ZooMS process, the sample becomes smaller. The use and choice of animations is structured around this progression, bringing the viewer closer in to each micro-world, zooming further and further in to emphasise the very small amounts required with this technique but to also try and create a more tangible experience of this unseen world. All the animations are made in 3D (with Blender 2.76) and while each animation has its own character the overall stylistic aim was to was to create a clearly computer-generated and unrealistic appearance, something very simple that would allow for a similar style or thread to be shared across the different sequences. Each animation also has its own requirements to take into consideration that are discussed in more detail later.
Hopefully the posts will be of interest, either from a scientific or an artistic view point. And if you’re a heritage professional with some parchment get in touch with York Uni BioArCh and have it sampled. The more samples there are, the better it gets!
A few excerpts from the making of a short animation showing the triboelectric effect. Part of a longer information video being produced for the University of York.
For more info on their clever techniques and findings, try these links.
The animation was made using Blender 2.7 and the sound and video assembled with various adobe packages.
a project to be de-mothballed
One of the greatest delights of an english summer are the rich and varied plants and blossoms that spring forth from every nook and cranny. Cheif among them are those that can be worked in to some form of culinary delight, be it the humble nettle or the fragrant rose.
Today’s concoction is Honeysuckle Delight – using a honeysuckle infusion as the base for some home made turkish delight.
An early draft of collage combining interesting elements of York running in a section through the whole city and out the other end. Begun in 2012, but on ice ever since.
Of cuts from a picture recently made as a gift – it is the view over one of Copenhagen’s lakes in winter, a layer of ice sits across it, pierced by the feet of geese.
Home made frames for home made photographs.
I have had the good fortune to receive 2 very fine, uk made knives from the skilled and friendly Will Ferraby.
Check him out at http://www.ferrabyknives.co.uk but be ready for a long wait, he’s quite popular!