By pure chance I caught BBC's digital technology program 'Click' (Freeview Ch80) this afternoon. The main topic was 3D recording of buildings and structural features by deriving the 3D co-ordinate information from multiple 2D images via some very heavy computing. Obviously this is not going to be as good as the results from the proper 3D imaging kit as used by EH, CADW etc. but it looked pretty good. I would strongly suggest that anyone who has ambitions or even dreams of doing this should go out of their way to catch one of the several repeats of this program, or catch it on the IPlayer assuming it ends up there.

Obviously the program was looking to take the photos on a mobile phone and access the software from the phone but I guess photos can be transferred to the phone from other better cameras. The biggest problem could be if the software has any built-in expectation as to the nature of the images it is processing but no harm in trying.

What was unclear was if you could get the resulting 3D point map back as data, or just the representations derived from it.

EDIT: From the URL kindly provided by JR in the next post the date is correct so it probably is the right one.

---

I think this is the one. "http://www.bbc.co.uk/iplayer/episode/b03bvxq5/Click_21_09_2013/"

---

Although I can't watch programs on th Iplayer in Germany but I assume that they were talking about software like Bundler, VisualSFM and Photoscan, all of which aren't really that new and are around for up to 4 years now. They work well surface and underground. If you like experimenting go for the free stuff, otherwise the commercial software will be a better choice for you.

Scale is the biggest issue with most software but you could overcome this with Meshlab. Otherwise you'll need the really expensive commercial software which is used for surveying.

You can export points, mesh and rendered surfacein various file formats.

This is a sample from 3 photos that I did earlier:

🔗Personal-Album-1711-Image-89999[linkphoto]Personal-Album-1711-Image-89999[/linkphoto][/link]

Have a look at Scenect and similar developments for Kinect and Xtion as well.

Digit

this is something we do as a company (Northern Archaeological Associates); and it is possible to geo-reference points within the photos on the surface, though obviously underground becomes technically more difficult. You are correct on the heavy computing - we generally have to run the program overnight, or for very big sites, over the weekend, but it does work. It is possible to output the data in different forms, perhaps the most usable are as 3D PDFs.

As to it being 'not as good as 3D imaging kit' (I presume you are referring to laser scanning or LIDAR), it is actually pretty good...we did some terrains with a kite-mounted camera and produced a point-cloud of similar definition to that of a LIDAR-derived model of the same area (and for a fraction of the cost).

The key issue with this software is the photos taken - you need an acute awareness of how many photos are needed and at what angles, or it won't work. The more complex the surface, the more photos you need...