Aerial Photos

A set of aerial photos taken at the end of our season using our two aerial platforms:

Site-wide 3D models...

Now that we are approaching the end of our excavation season, we are able to start doing more aerial photography of the excavation areas. These photos can, of course, be used to build 3D models.

Here is a model of Marj Rabba, Area CC, using final photographs taken with our quadcopter. 

And here is a 3d model of the entire main excavation area. This was built using over 100 photos taken with a GoPro Hero 2 camera in our r/c plane. The model is not fantastically textured, but it looks pretty good and really shows the area right around our excavation:

More software...

Here is a screen capture from my first test of Photoscan, a software package for photogrammetry that lets you export models and DEMs for use in a GIS. This is some pretty cool stuff. The main drawback of photoscan is that, unlike the other software that I have preferred (hypr3d, photo synth, and 123dcatch), this is not distributed "cloud" computing. As such it tends to hog the processor for hours at a time. I first tried to do a high res model with 190 photos. The first of 3 steps took 12 hours.

This is a top down look of area AA at the Chalcolithic site of Marj Rabba:

Stereoscopic Images - Maitlands Mesa

Ive just been playing with Stereoscan to see how well agisoft products works at creating 3d models. Stereoscan can make suprisingly awesome 3d models using only 2 photos, but what it is really for is creating traditional stereoscopic images. Below is a stereo pair of 2 photos taken with our fishing pole PAP setup at Maitlands Mesa, Jordan.

Stereoscopic images are made by taking 2 photos of an object from slightly different angles, approximating the different perspectives that your two eyes give you naturally. If you focus on the two images correctly, you can view the original scene or object in 3D. Here is a link to some more in depth directions for how to view these.

This is structure MTSS11, which we excavated extensively, but these photos were taken before excavations began:

Here is a nice 3-d model of a stone lined pit at Marj Rabba. This was excavated in 2010 and 2011 but remains in fairly good condition. This pit is cutting two of the enigmatic round features which make up the bulk of the architecture in this area of the site:

Experimenting with 123d catch for photogrammetry rather than hypr3d.

This is a short video fly-through of a structure from the site of Maitlands Mesa, Jordan:

Actual results from the field...

I just got back to Amman after 2 weeks of digging, mapping and shooting at a site in Jordan. This was my first opportunity to apply some of this stuff on a project I was a part of.

The landscape was spectacular:

 

YMR descending the mesa to our campsite at the end of the day, with a view out into the wadi

and at least some of the photogrammetry stuff I shot using PAP (Pole Aerial Photography) is going to work. We shot a lot of low elevation PAP using a longer 8m fishing pole I bought just before heading out to the field. It looks like those shots (from 8m up) are great for publication quality vertical images of archeological features and excavation units:

 

But low angle shots really work best for photogrammetry. The following two models of structures we excavated were built using, primarily, photos taken from less than 3m above the ground: 

Storehouse at Be'er Sheva

This post will be some of the last practice 3-d photogrammetry I do before the field seasons start. My process for building these is getting pretty refined. This is the most successful model Ive made yet. Im now using a short "painter's pole" to take sets of near-vertical overlapping images. Im using our point-and-shoot canon which works as well for these models as my big SLR, and is easier to work with because the images are smaller. The camera is running an "intervelometer" script that takes a picture every (in this case) 3 seconds. I look like a doofus taking the pictures:

But the results work better than taking photos from eye level. Structure-from-motion programs seem to work best with high angle shots that dont include the horizon. This model of a store house at Be'er Sheva was constructed from 150 overlapping photos. Most were shot as close to vertical as possible, a few shots of the standing column bases included some shots closer to 45 degrees (as always, these look better in full screen mode, and you should click and drag the models around):

And this is a detail of a stone lined basin at Arad:

The complete system!

Our archaeology photography platform is ready to go:

We are using an FMS Skytrainer -

 carrying either a GoPro Hero 2 HD or a small point and shoot Canon. Both are capable of shooting in "intervelometer" mode (the canon uses CHDK). 

Here you can see the camera just peaking out from the belly of the plane and secured by a series of rubber bands:

The plane has been tested several times with the camera mounted in the belly for vertical shots as well as on top for horizontal shooting: 

Now that the plane is fully functional we took it out to test it at actual archaeological sites. The plane can shoot images for two different purposes. The first is oblique shots of a site that show the surrounding landscape and give a good overview of the site and its surroundings. We shot like this at Tel Be'er Sheva by placing the GoPro Hero in its widest, nearly fisheye, setting. Then, by flying next to the tel and banking to get a nearly 45 degree view, we got some good oblique shots. This is the kind of thing you might use for a presentation about a site: 

The camera is set to take an 11mp photo every second. This means that for even a short flight of 4 minutes or so, we generate hundreds of photos, including tons of pictures after the shutter has been pressed but before the plane actually takes off (so lots of pictures of pebbles from 6" off the ground). This means there are lots of images to sift through to find useable ones. Here you can see me, NB, and JC from quite high up. Which is cool, but the archaeology is nowhere in sight:

The other kind of photo is vertical images which do not suffer from ultra-wide angle distortion and can be used to show the spatial relationship among areas of a site. These kind of photos are critical for good publication and recording of archaeological sites. Here you can see some vertical photos of Tel Burna taken with the GoPro Hero on a less extreme FOV (field of view) setting (though still not devoid of distortion:

NB, JC, and I are just visible at the right edge

Aerial images like this can also be used to construct 3d models of sites and landscapes. Here is a partial reconstruction of Tel Burna using a set of images that includes the previous 2 shots. Unfortunately, the camera ceased recording about 1/2 way through this flight, so there was not enough imagery to reconstruct more of the tel or to connect the higher area:

with the lower area:

Results of shots from our own excavations using this platform will be posted here as they are available. 

WOW - RCAP and 3d Photogrammetry!

This is the first result of 3d photogrammetry from images taken with our aerial photo platform. This looks phenomenal.

We successfully tested the Galilee Prehistory Projects aerial platform using the go pro mounted horizontally. Photos came out quite well:

PAP (Pole Aerial Photography) Test

So I bought a 2m pole, and am experimenting with shooting 3d photogrammetry with it. This is made from photos taken with a canon running an intervelometer CHDK script, flowers don't work at all for 3d photogrammetry, and I failed to get an angle that includes the near side of the fountain, but it will work, and you can't beat a 5 dollar aerial photo system:

Grain Silo at Tel Dan

Larger 3d Model - The middle bronze gate at Dan

This came out remarkably well for being hard to photograph (limited angles) and in a very mixed light.

Architectural detail at Tel Dan

Architectural Detail of in-situ and reconstructed sections of the Bemah at Tel Dan

Reconstruction of an olive press at Hazor

Small altar at the Tel Dan High Place

Hazor Pillar Base

Basalt pillar base at Tel Hazor:

Tel Hazor

Detail of the late bronze age palace at Tel Hazor:

Test of the new aerial platform for this summer

Kursi - 3d reconstruction of a room at the monastery

Floor is incomplete (I ran out of space on my memory card as I was finishing), but it still looks pretty good.

N.b. you should make this full screen and then zoom into the middle of the room and look around.

Incomplete 3d mesh of a room, but looks pretty good.

This was shot in a rush from relatively few angles, but looks OK, despite there being big holes in coverage.

This is a 3d mesh of a site built using pictures that were not collected with 3d modeling in mind. The mesh looks OK for pictures that do not go 360 degrees around the site and have no low angle shots (so provide little relief). This will look much better with more complete collection.

Has some holes, but this works pretty well...

More testing...

I took pictures to try to make a 3d mesh of the column in the middle here, surprisingly, the surrounding columns also were digitized, though with very little detail. They could be edited out, but its not worth the trouble for this test.

Testing 3d Reconstruction of archeological features