Loading ...
Sorry, an error occurred while loading the content.

52274Re: [PanoToolsNG] Re: OT - creating spatial models from 3 panoramas

Expand Messages
  • Bjørn K Nilssen
    Nov 8, 2011
    • 0 Attachment
      På Tue, 08 Nov 2011 16:49:20 +0100, skrev enridp <enridp@...>:

      > I was looking for a tutorial using tgi3D with panoramas but I couldn't find anything.
      > I don't have too much experience in 3D, what is the workflow for making a 3D environment using a panorama?
      > Do we need to import the 6 cube faces and make a 3D cube in sketchup? why we need 3 panoramas and how can we work with them together? I'm a bit lost :(

      Tgi3d Photoscan is not using panoramas/equrects as input, but normal rectilinear photos. Such rectilinear photos can be extracted from the pano in PTGui, DevalVR etc.
      PhotoScan calibrates those photos, undistort them as needed, and exports them to SketchUp as a set of construction points from which you can build your model using the SU and PhotoScan tools.
      The calibration can be done with at least 2 photos, shot from different locations, but showing at least 8 points that are visible in both photos.
      It is usually better with more shots though.
      You add a marker in one photo, and then add another marker at the same spot (in 3D space).
      These two (or more if you have more than 2 photos where it is visible) markers are "paired", telling PhotoScan that they share the same 3D position.
      Pretty much like adding control points manually in PTgui, just remember to never use any highlights or other similar points, because they move between photos shot from different positions.
      When you have enough pairs (or sets) the calibration is done.
      The program now knows the actual position of the 3D points, as well as the camera positions.
      It can be improved by adding more points (and/or photos).
      With good photos and a good reference measurement you could easily model a normal sized living room with a precision of less than 5mm from those photos.
      You could also mark a 3D point as origin (0,0,0), set a known distance between 2 points to set the scale, and align the coordinate system by marking 2 pairs of 3D points.
      Next is export to SU format (or other), and then you get each photo as backdrop, making it a lot easier to create the model.
      As opposed to IMageModeler and PhotoModeler you are not limited to only model by snapping to the calibrated points.
      PhotoScan has a very powerful tool that lets you draw a line or curve in one photo, and then view lock it.
      When you move to the next photo that line/curve may appear to be floating in the air, but you can now drag each endpoint to the corresponding point in that photo.
      Now it has been correctly placed in 3D space.
      Because it was locked in the first photo it will not move at all from that viewpoint, and as it is now looking right in both (or more) photos it implies that its location is correct. If you used a pair of the imported construction points for snapping to with the line tool in SU it would of course already be located correctly by adding it in only one photo, because the endpoints were already correctly located.
      This viewlocking may sound like a simple little tool, but it is actually immensely powerful, not only for photo modeling.
      It allows you to model anything in the photos, regardless if you have any calibrated points on them or not.
      You can't do that in ImageModeler or PhotoModeler...

      > --- In PanoToolsNG@yahoogroups.com, Bjørn K Nilssen <bk@...> wrote:
      >> På Fri, 04 Nov 2011 13:07:38 +0100, skrev enridp <enridp@...>:
      >> > I found these DVDs if anyone is interested:
      >> > http://www.thegnomonworkshop.com/store/category/58/Greg-Downing
      >> >
      >> > But it would be nice to have an updated tutorial, I guess today there are many advances in hardware and software which can help to make the same thing with less effort and in less time.
      >> Not really. Quite the opposite actually ;)
      >> Unless you're the owner of an Autodesk product like Max or Maya?
      >> Then you can get access to (buy?) ImageModeler, which is what Downing was using in those tuts.
      >> I've used ImageModeler for many years, but when it (Realviz) was bought by Autodesk things started to happen. For a while it was available freely, but now it is not possible to buy unless yo're already a paying customer, which I'm not.
      >> My version cannot use panos as source, but the Autodesk version I tested with that feature was extremely buggy and crashed all the time. At the same time they had removed some of the export formats! I never upgraded, and then the doors were closed.
      >> Photomodeler is another similar tool.
      >> Canoma is another oldtime favourite, swallowed by Adobe, and never seen again thereafter, unfortunately. It was very inspired by Debevec and that Campanile movie, and very fun to use :).
      >> My favourite today is SketchUp with the PhotoScan plugin from tgi3d.
      >> You'll find tons of tutorials on YouTube.
      >> Erik Leeman is a brand new SketchUp user (but long time 3D and pano experience), and here are some of his pano/cubefaces->3D pictures on Flickr:
      >> http://www.flickr.com/photos/erik-nl/6301352751/in/photostream/
      >> Looks very good :)
      >> All those mentioned above are polygonal/imagemapping photogrammetry tools, but there are also a type called point-cloud tools. Laser scanning and software/photo-based scanners typically end up with lots and lots of data, and usually also lots of holes.
      >> > --- In PanoToolsNG@yahoogroups.com, "enridp" <enridp@> wrote:
      >> >>
      >> >> Wow ! this is incredible, and it's from 10 years ago (I think):
      >> >> http://www.gregdowning.com/pimp/index.htm
      >> >>
      >> >> Also this has a really great transition between 3D and Photo:
      >> >> http://www.gregdowning.com/3dqtvr/index.html
      >> >>
      >> >> Does somebody know a detailed tutorial for making things like that?
      >> >>
      >> >>
      >> >> --- In PanoToolsNG@yahoogroups.com, "Hans" <hans@> wrote:
      >> >> >
      >> >> >
      >> >> >
      >> >> > --- In PanoToolsNG@yahoogroups.com, "jeffreycb2000" <jeffreycb@> wrote:
      >> >> > >
      >> >> > > Many many years ago there was a post to a web site that was researching how to recreate a plaza from 3 spherical panoramas from differing positions. The output model re-created the plaza space and had the photographic imagery attached to the various surfaces and objects within the model. This allowed the viewer to then walk through the model anywhere (within the limits of the panorama coverage).
      >> >> > > Testing peoples memory - but would anyone know of the site or something similar.
      >> >> > > I've a friend who is investigating from the laser scanning and pano overlay side and I thought this would be interesting to show them.
      >> >> >
      >> >> > I guess you talk about Greg Downings Notre Dam
      >> >> > http://www.gregdowning.com/3dqtvr/index.html
      >> >> > http://www.vrmag.org/vartist/guest_artist/A_CONVERSATION_WITH_GREG_DOWNING_TRAVEL_PANORAMAS_3D_TECHNOLOGY_AND_HDRI.html
      >> >> >
      >> >> > Or the original Campanile movie, by Paul Debevec
      >> >> > http://ict.debevec.org/~debevec/Campanile/
      >> >> >
      >> >> > Hans

      Bjørn K Nilssen - bk@... - 3D and panoramas
    • Show all 26 messages in this topic