Zyra International NET //// Archaeology //// Computing //// Site Index in Pieces

Putting the Pieces back together in Archaeology

Archaeologists often dig up things which are in pieces. It would be nice to be able to put the pieces back together to reconstruct the original. This is not easy.

However, I have an idea about it which could help. Supposing an archaeological site is excavated and there are a great many pieces, it might be possible to use the recent advancements in computing to piece together "jigsaw puzzles" which had previously been impractical. Even if the remains consist of many thousands of fragments, like gravel, they could be pieced back together.

I've seen documentaries about archaeological sites where there is rubble about, and I wonder if the pieces might fit together to reconstruct the edifices which long ago existed. I think this is especially likely when the pieces are of complex 3D surface and made of a rigid material.

Example 1: Gramophone Record

Supposing a while ago someone smashed a 78 RPM gramophone record against a wall and all of the pieces fell into compost on the garden. In a later period, archaeologists excavate the ground and rescue all of the pieces. Here's how the record could be reconstructed and played:

1. The entire area is put through a sieve and all of the fragments are collected.

2. The fragments are washed and then put through a three-dimensional scanning machine that creates a "virtual" version of each piece in high-res quality in a computer. The physical pieces are numbered and stored separately.

3. The computer does "jigsaw puzzle solving" and finds which pieces fit with each-other and how.

4. In the virtual environment, the virtual pieces are assembled into a virtual complete record.

5. The result is played on a virtual gramophone player inside the computer, the output of which is a soundtrack.

6. The soundtrack is played in the physical world.

Update: Something a little like this in some ways (using 3D imaging to recover a sound from an 1888 phonograph cylinder that had been bent) can be seen at www.cbsnews.com/stories/2011/07/06/scitech/main20077308.shtml and www.nps.gov/edis/photosmultimedia/talking-doll-record-hear-the-recording.htm and http://theconservativetreehouse.wordpress.com/2011/07/19/thomas-edisons-doll-talks-after-123-years-of-silence/ and one where you can actually download the sound: www.archive.org/details/EdisonsTalkingDollOf1890

Of course with a gramophone record there are various clever shortcuts that can be made, for example taking into consideration the fact that the record is a flat disc, and the fact that the grooves have a radius which is a pretty good indicator of their distance from the centre. However, I believe that the idea of putting the pieces back together from bits unearthed can be generalised and can be applied to pieces discovered.

The virtual piecing together can be done for pottery, glass, rigid stone, and other durable rigid solid items which form fragments with clear edges.

I was inspired to write about this after I saw a documentary about Israeli archaeologists who had inherited several tons of mixed-up rubble which Muslim archaeologists had dumped after excavating stuff under the Temple Mount. Admittedly they were finding interesting fragments of history, but I wondered if the idea of extra precision in archaeology could be carried a stage further. Maybe the smashed pieces of foundation stone could be pieced together, in computer. In fact, maybe a general purpose algorithm could be found that could reassemble, in a virtual environment, anything that had previously been smashed.

To take a silly example, supposing someone was intent on doing the dish-washing, but instead of putting the crockery in the dishwasher, they put it in the tumble dryer. Can the fragments of smashed plates, cups, etc, ever be put back together? My belief is: Yes, they can. Not everyone agrees with this, and it has been pointed out that in fact such problems are extremely difficult, and that for larger numbers of pieces, it is increasingly impractical with increasing numbers.

However, I have noticed that computers get more powerful year-on-year, and as well as that, people manage to write programs that are cleverer. There are often neat techniques which make previously near-impossible tasks relatively easy. For example...

* I offloaded a sequence of 202 wildlife photographs from a digital camera. Unfortunately something went wrong and there were "Segmentation Fault" errors coming up. In the end, I had got the 202 images, but I also had about thirty "temporary" tmp*** files.

* I wondered if all of these additional tmp files were identical copies of other images, or if there were any additional ones.

* Conventional wisdom on this says that I should look at all of the images and see if any of the tmp files are the same as any of the 202 original images. However, this would require about six thousand comparisons.

* Instead, in Linux command-line I did:

cksum * | sort | less

* This produced a list of all the checksums of all of the files and listed them in order of checksums alphabetically. I then looked through them as a list and for every tmp file I looked to see if it did had a unique checksum or not (by consideration of its neighbours). This required sixty comparisons in text format, and was accomplished in about two minutes. There were no unmatched files. All of the tmp files had a perfect match with an original image. However, if the converse had been true, ie that one or more tmp files had a unique checksum, then it would have been an image not included in the original set.

Another example of difficult things made easy:

* Supposing someone remembers a tune, but they don't know what the name of the tune is.

* As most people can't reproduce the tune with great accuracy, this makes the identification of the tune rather difficult.

* A website was set up that required the person merely to identify each note well enough to say whether it was of higher or lower pitch than the note before it. Interestingly, this works well enough to identify a tune within a few notes.

Back to the practical archaeology, I believe that the piecing-together of 3D objects can be accomplished, even for large numbers of pieces. If I am right, then the fullness of time will prove it. If I am wrong, then there will be some inherent difficulty to which I am oblivious. It will be interesting to find out whether I am right or wrong about it!