none
Body measurements? RRS feed

  • Question

  • I wondered would it be possible to use the SDK Kinect a human body to measure?
    Not only in front but the whole body. For gym etc.
    Any ideas?
    Cogito Ergo Sum
    Monday, July 18, 2011 10:57 AM

Answers

  • It's certainly possible:

    http://www.exactmetrology.com/breuckmann_bodyscan.htm

    Practical is a whole differant matter. It would be hard enough with a rigid body and an articulated body like a human with skin deformation added in makes for quite a challenge. That's a career, not a hobby. Piecing together multiple meshs with a rigid body is no easy task. Humans move and their skin deforms as they move. Fat people deform a whole lot as do someone like a body builder, but even middle of the road people do, i.e. twisting the torso. They breath, they figit, they shift their weight around. It would be a whole lot easier to scan them if you killed them first, but that would be bad for business.


    Easier would be just do a profile, i.e. use the player segmentation. Take front, back, left, right and capture the depth data while you're at it. You can then easily show them before and after. That would be, pretty much, like showing them photos. So you can do it professionally without a whole lot of effort. Capture the depth and skeleton data as you do it and you start accumulating test data. Just getting something useable, practical and profession using four independant 3D views is an accomplishment. It's sort of like a body mold. It has some challenge adjusting poses so you have to start attaching the skin to the bones. That way you can move the feet slight in and out and such. Once you actually attach the skin to the bones then it becomes much easier to match up the meshs.
    Tuesday, July 19, 2011 6:14 PM

All replies

  • Well, in theory yes, but the Kinect can only perceive what it sees, which is one side of the body at a time. I guess the person could be made to turn in front of camera, or you could have multiple kinects positioned at different angles around body being measured to do it without need for turning human around.

    What kind of user experience are you looking for in your application?

    Eddy


    I'm here to help
    Monday, July 18, 2011 9:04 PM
  • Well, in theory yes, but the Kinect can only perceive what it sees, which is one side of the body at a time. I guess the person could be made to turn in front of camera, or you could have multiple kinects positioned at different angles around body being measured to do it without need for turning human around.

    What kind of user experience are you looking for in your application?

    Eddy


    I'm here to help


    Eddy, thanks for your response,
    My daughter and husband have a gym with heat cabins for quick weight loss.
    In the beginning se measure and weigh them customers. Now I thought, you can use the MS SDK Kinect and scan the shape of the body?
    I can create a platform that is turning 360gr, no problem.
    The length of the customers measurements would  also be nice. ( maybe a raster on the wall?)
    Scales in the platform. Etc. ....

    Regards Henk.
    ps: sorry about my English


    Cogito Ergo Sum
    Tuesday, July 19, 2011 8:08 AM
  • It's certainly possible:

    http://www.exactmetrology.com/breuckmann_bodyscan.htm

    Practical is a whole differant matter. It would be hard enough with a rigid body and an articulated body like a human with skin deformation added in makes for quite a challenge. That's a career, not a hobby. Piecing together multiple meshs with a rigid body is no easy task. Humans move and their skin deforms as they move. Fat people deform a whole lot as do someone like a body builder, but even middle of the road people do, i.e. twisting the torso. They breath, they figit, they shift their weight around. It would be a whole lot easier to scan them if you killed them first, but that would be bad for business.


    Easier would be just do a profile, i.e. use the player segmentation. Take front, back, left, right and capture the depth data while you're at it. You can then easily show them before and after. That would be, pretty much, like showing them photos. So you can do it professionally without a whole lot of effort. Capture the depth and skeleton data as you do it and you start accumulating test data. Just getting something useable, practical and profession using four independant 3D views is an accomplishment. It's sort of like a body mold. It has some challenge adjusting poses so you have to start attaching the skin to the bones. That way you can move the feet slight in and out and such. Once you actually attach the skin to the bones then it becomes much easier to match up the meshs.
    Tuesday, July 19, 2011 6:14 PM