none
kinect v2 / fusion / scanning cars RRS feed

  • Question

  • hi, 

    im looking for the best solution to use the kinect v2 to scan entire cars or car parts in a controlled indoor environment. My goal is to get a 'clean' mesh (fusion?) with 1-2mm accuracy and color information and use it a 3d modeling environment after.

    I could use multiple kinects to cover the entire surface or a single sensor moving automated on a track but due to the size of the scanning room the sensor cannot be further away than 1.5m.

    thanks, 

    Peter

    peter@thinkmatik.com

      

    Monday, August 31, 2015 9:20 PM

All replies

  • Any suggestions?
    Friday, September 4, 2015 1:31 AM
  • You can try KinectFusion (the sdk version or others online like Kinfu LargeScale http://pointclouds.org/documentation/tutorials/using_kinfu_large_scale.php or voxel hashing http://www.graphics.stanford.edu/~niessner/niessner2013hashing.html) but I doubt you will get the results you want. 

    Cars are pretty complex shapes (overhangs, thin structures), with lots of different textures + colours, and the colour integration of KinectFusion is not great as you get shadows, bad colour integration etc.

    For this task where the object is static and time is not an issue I would use something like

    http://www.123dapp.com/catch

    http://insight3d.sourceforge.net/

    http://www.acute3d.com/smart3dcapture/

    and use a very good DSLR (or multiple cameras)

    Friday, September 4, 2015 7:06 AM