Thursday 16 February 2012

Kinect, real time 3D

This week we are doing some lessons in Lund University and yesterday evening we had the opportunity to teach a couple of hours for the Knut student association.
To avoid repeating the same argoument we did during the course, we tried to play a little bit with Kinect and test RGB Demo, a toolkit released under LGPL by Nicolas Burrus.
The results were very good, expecially if you think that RGB Demo has an application to record 3D data in real time. The quality is not comperable (by now) to what you can reach with SfM/IBM techniques, but the system is very much faster. We think that it could have some utility archaeology, so probably we will package the software for the next release of ArcheOS (theodoric).
In the picture below you can see a screenshot of the reconstruction of our room in Lund (Sparta guesthouse), which is our fist test with Kinect. Soon we will update other considerations on the software...


4 comments:

  1. Kinect are fantastic. They join in the same device a color camera and a infrared laser, to record distances. With the correct software, it can really work like an affordable laser scan (with texture photo mapping). Apparently the infrared sensor range is very limited and work poorly under direct sunlight.

    Probably can be very useful to record caves, tunnel systems, or indoor archaeological excavations. Maybe in cave surveying it work better than the SfM/IBM approach. SfM/IBM requires very good lighting to obtain perfect photos - a difficult thing to obtain underground - but probably Kinect laser can even work without light (just guessing). Ergo, creating a (non colored/texturized) 3D model.

    See also:
    http://idav.ucdavis.edu/~okreylos/ResDev/Kinect/

    rgaidao

    ReplyDelete
  2. hi Ricardo,
    yes this is more or less what we are planning to do: testing Kinect in caves and other dark places. By now we get very good results also with SfM and IBM, for example working indoor with very low lights (in a church). In this case we just use a tripod for the camera and the 3D model was very accurate. Anyway by now we did not get such an accurate 3d model with Kinect, but the good things are: real time 3D recording and the possibility to work in dark places (infrared camera). The fun with Kinect is just started...

    ReplyDelete
  3. I eagerly expect the results!

    P.S. - Just received this info: Using a inertial sensor to survey the subterranean roman aqueduct of Tarragona (Spain).

    “Amb un sensor inercial tindrem un mapa acurat del 'cuniculus' de Tarragona”

    http://www.icac.net/index.php?option=com_content&view=article&id=545%3Aqel-cuniculusq&catid=4%3Atauler-anuncis&Itemid=4&lang=ca

    rgaidao

    ReplyDelete
  4. Hi,
    I will read the article very soon (now I have to go to the lesson), but I guess they used something like an accelerometer (maybe with an electronic compass). Can it be? In this case would be possible to assembly something like this with part of the components of the UAVP or of the others quadcopters projects (maybe also simply hacking a Nintendo Wii, ore developing an app in Android for a smartphone which has an accelerometer). It is really a nice application, they had a very good idea!
    Ciao

    ReplyDelete

BlogItalia - La directory italiana dei blog Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 International License.