Archaeological Planning using Photoscan- without GC

[Hughsnews]

Hi All,
I, like a number of folks here, am an archaeologist and I have used PS to create models of archaeological remains- although I am fairly new. I am after a bit of help with a workflow for something slightly different.

Here is the situation- on a couple of projects I have been on, I have created top down plans of archaeological features using photoscan and a total station to give me GCP. Basically, I create orthophotos which I trace off in adobe photoshop or illustrator. The plans end up looking like this(http://www.twoatlarge.com/illustrations/tower_1.jpg) nb: thats not my plan and just shows what an archaeological plan will look like if you don't know. I will sometimes make models without a total station but will mark on the stones a series of measurements ie I will put stickers on that are 1m apart to give me a scale.

I have been asked to look into using this sort of skill on things such as rescue excavations, where we have a day or two to dig an area and require a plan, but often not the time to do this properly. The big issue with this is that generally there are no total stations or anything to give me x,y,z coordinates. I can us a theodolite to get heights(z) though.

What I need to do is to produce an orthophoto that is a perfectly vertical/birds eye view of the building, wall , feature etc like the example above. What I am curious about is if anyone can suggest a way I can get this perfectly vertical orthophoto without just trying to guess what is vertical? With GCP points from a TS- it obviously knows what is flat and produces a nice orthophoto, I just cant work out how to do it without GCPs.
 I would then use stickers or markers at set measurements to provide scale so in that sense it's fine.
 I am using PS Professional.
If anyone can offer some advice, or if there are any archaeologists out there doing this, I would love to hear from you.
Thanks
Hugh

[Alexey Pasumansky]

Hello Hugh,

If you can place three or four markers corresponding to the points of an equal altitude (according to theodolite measurements), you can use them for Planar projection definition in the export orthophoto dialog. Similar to the project from the recent forum thread: http://www.agisoft.com/forum/index.php?topic=3586.0

For that I assume you'll need to measure the elevations of some points in the excavation site prior to the image acquisition, find out points with the same elevation and mark them somehow, so that you can easily place markers during processing in PhotoScan.

[ThomasVD]

If you just need the model to be perfectly vertical you could also take a simple object such as an L-shaped ruler or even a square piece of cardboard (it just needs to have two axis), use a water level to make sure the L-shaped ruler or the rigid piece of cardboard are lying 100% horizontal along both axis, ensure the object is included in the photogrammetry models and then set markers on the edges of the ruler or cardboard to define your plane in PhotoScan.

[bigben]

You might also consider calibrated scale bars for some situations to save time with measuring. e.g. http://culturalheritageimaging.org/What_We_Offer/Gear/Scale_Bars/index.html

[Geo]

Hello Hugh,

You may want to consider the following tools for your type of project.
1) use a small 30metre steel tape in a rolling winder (approx. cost $30) to create scale lines for your project.  Depending on the scope of your project, you could typically layout three control markers on a straight line with your tape.  Then create two other control markers by using the mid marker and each end marker by setting two markers on one side of the 3marker line.  Tape the distance from the end marker and the mid marker to the two point outward from the 3point line.  These are the distance observations for scale.

2) Use a second hand builders level (very inexpensive) and a surveyors rod to measure relative elevations of all five markers. These are the elevation difference observations.

3) use a hand held GPS (everyone has one right) to record averaged positions (N,E) for the five control markers.  These are the coordinate observations.

4) If you set out the three points on line properly carefully, then you have an angular observation being 180 degrees at the midpoint

5) Enter your observation data in the excellent free Gama Least Square adjustment program by Ales Cepek and team.  Use the appropriate standard deviations for the observations.  For example for the 180 angle observation use "0" seconds thereby fixing the straight line. Maybe use one centimeter for the distance and the elevation difference observations.  Use two to ten metres for the standard deviation of the autonomous GPS.  You will first run a separate vertical adjustment with the elevation difference observations and thereby obtain adjusted elevations and have a check on the quality of the observations.  Secondly you will run the horizontal adjustment.  The Gama least square adjustment will respect the weighting that you used and provide final real world coordinates and relative accuracies.  You could even add in forward and reverse compass azimuths with say a standard deviation of about 2 degrees, but not needed.  The adjustment will show you how good your autonomous GPS observations as are as well. I think typically around the 2 metre absolute positioning level and Gama will make the distance observation best fit the GPS positions of course.

6) When you add the control points in and run Agisoft, the result will of course show how well the control fit the photo alignment/mesh which provides another indication of the accuracy of the input scale measurements. 

This may seem overly complex to do, but once you have the abstract format understood, it is very quick to do for these small networks and very rigorous way t do it.

Gnu Gama can be found here:  http://www.gnu.org/software/gama/
I am not sure whether this user interface for Gama up to date and working
http://roci.sourceforge.net/

I created a little resource site about ten or fifteen years ago that has some sample input file and resource info on using Gama
http://vrmapper.com/gama/gama.html

There should be sample horizontal and vertical input files that you can simply use and fill in your own data.  There is an instruction on how to install and run Gama as well.  It is a very good/powerful and easy program to use once you complete your first project.  I think it is ideal for use on small archaeological or similar sites.

I hope this helps,
Jim

[Hughsnews]

Hi All,
Thanks for the advice. So I have been tinkering around with a few models I have. They are of archaeological trenches, and have pegs, which I know the distance between. These pegs have been placed in with a total station, meaning despite changes in elevation, they are set distances apart ie 5m if you were looking at it from a orthoview. I do not have the X-Y-Z coordinates though.

What I did was added each peg as a marker with no xyz values and then created 5 scales and entered the distances I knew were between them. They circled the trench on atleast 3 sides. Ie think of a square and I knew the distance of the three sides. On one model, these were all at different height levels.

When I produced the model- I noticed that not the FRONT view but the Back view seemed to give a pretty vertical orthophoto. I saved it as an ortho and then in photoshop overlayed two different plans of walls in the trench and both of them aligned pretty much perfectly.

Now I am fairly new to this so if someone can explain how this happened-ie how Photoscan worked that out? Is this, as I knew the horizontal distances between each marker that PS could use those to work out what was flat/verticle?I would have thought once the model was created and became 3d, that measurement of 5m would be very wrong as the pegs are vertically at different heights. Basically, many of you said something similar below ie Thomaxs,Alexey, Geo.
I imagine if I made a model with pegs that were 5m apart, but that was just measured by tape and wasn't all at the same height- it would be askew.

So in order to replicate those without total station pegs, as many of you have said, if I was to set up several markers which are horizontally the same height, and whose distance apart I knew, it would produce a completely vertical. Say a few flat scale bars 1m wide with markers on either end, that I use a spirit level to make sure is perfectly flat. I imagine if I used two or three of those, they all would need to be the same height?

As Thomas said, if I used an L shape piece, say 1m by 1m, with 3 points(one on each end and one in join) that was nice and flat, even on a large model of say a building that was 20m by 20m, as long as its in model, I could use it to make sure it was flat even though its quite small? Or is the further apart the points, the better? Does that make sense?

Sorry, for no doubt stupid questions, I am just trying to wrap my mind around it. Would hate to do something and find that I messed it up and that the archaeoloy was then destroyed and I couldn't check it.
Thanks so much for all your wonderful help.

[frank.stremke]

hi,
just use the theodolite to survey in your control points by trigonometric survey or by polar survey using the theodolith and tape meter
frank