3D Segmentation using ReconstructMe

Our (PROFACTOR) interest in ReconstructMe goes beyond reconstructing 3D models from raw sensor data. The video below shows how to use ReconstructMe to do a stable foreground/background segmentation in 3D. A technique that is often required in 3D vision for pre-processing purposes.

With ReconstructMe you can generate background knowledge by moving the sensor around and thus building a 3D model of the environment. Once you switch from build to track only, ReconstructMe can robustely detect foreground by comparing the environment with the current frame.

This technique can be used for various applications

  • Monitoring robotic workcells for human safety aspects.
  • Intelligent reconstruction of process relevant objects only. We definitely do a video on this one.

to name a few.

21 thoughts on “3D Segmentation using ReconstructMe

  1. Bartlomiej

    Hi guys. I love your progress. One question: Is there a possibility to take part of the testing? Read trough your google groups news.
    Maybe you will make a BETA 3.0 phase. I apply for that :)

    Reply
    1. Christoph Heindl Post author

      Bartlomiej, we will release the public beta 03/2012. Then, everyone is free to test and reconstruct :)

      Reply
  2. Jaco

    Very good indeed.. What would happen if you spun the chair with the tube on it. When the chair spins the tube will be hidden from view while the back of the chair is towards the camera. Will it loose track of the tube then?
    Just a thought. :)

    Reply
    1. Christoph Heindl Post author

      It woudn’t loose track, but the chair would temporarily go green (until it has reached the ‘background’ position again). Best,
      Christoph

      Reply
  3. Peter

    Interesting feature on 3D segmentation. Is it possible for ReconstructMe to detect a moving object from a stationary (kinect) sensor. It would be similar to the application of motion detection from a (2D) video. Look forward to get your module if you have such a 3D motion detector.
    Cheers
    Peter

    Reply
    1. Christoph Heindl Post author

      Could you elaborate your thoughts? In this video the sensor detects a moving object from a stationary and moving sensor perspective.

      Reply
    1. Christoph Heindl Post author

      We will have a meeting today, discussing the features for the final BETA phase (which willl also be the version that is publicly released). We will discuss limiting features (maximum reconstruction time, maximum volume) and removing some yet instable features.

      The upper bound are of course hardware limitations.

      Best,
      Christoph

      Reply
    1. Christoph Heindl Post author

      That’s hard to tell, but it works well with a nvidia 560 GTX (~150 Euros), or AMD 6850 (~130 Euros).

      Reply
        1. Christoph Heindl Post author

          We are using only one device. However, if you bridge multiple cards so they appear as one device it will help. In general I’d recommend a more powerful card though.

          Reply
  4. Jonas cvFun

    Hello, wawww I love so much your work guys.

    This latest video is so interesting about extinding reconstructme to dynamic scene scenario. Does this option available in your demo? if yes, what is the line command to run it?

    Cheers

    Reply
    1. Christoph Heindl Post author

      Yes it is enabled, no switch needed, just do normal realtime scanning and try to move things around.

      Thanks for your feedback!

      Reply

Leave a Reply

Your email address will not be published. Required fields are marked *