next topic »

[on:]

Okay this is my problem: I am attempting to make a device that will give me a position (or change in position, either one) like a GPS would, and a rotation but the gyros cover that already, this would be installed on the top of a xbox Kinect which would be connected to my computer (I am trying to make a 3D object scanner) so I need it to give me something along the lines of "You are x,y,z away from where you started" a few times a second (obviously just the numbers). I am not using this over long distances (I would just be moving at most five feet over a couple of seconds) but a GPS is not accurate enough for this, at least not any that I have seen. I need it to be at the very least within two inches for accuracy.
Assuming the accelerometer and gyro are accurate I know the math needed to be done to get the position. Also I know that the longer you would use this the more error you would have but like I said I am only using it for short spans at a time.

So my first question is: Is this feasible? Would it be accurate enough? Or would there just be to much noise? (Maybe someone has done this before and can tell me how stupid I am for trying to do this...)

My second is just are there any other suggestions for how to get a position in "3D space", i.e a more accurate GPS, or some other method?

I've seen a video where someone has done this, although I don't know what method he used and his was a little less mobile than I'm hopping mine will be.
Any help would be awesome, thanks guys!

[Reply #1 on:]

So my first question is: Is this feasible? Would it be accurate enough? Or would there just be to much noise?

I think there would be too much noise which will result in drift in position.

You are probbly best exploring an ultra sonic solution with three or four external transducers pinging a receiver on your mobile device.

[Reply #2 on:]

If you move slowly enough your 3D scans should be highly correlated.  Perhaps some math would be enough to determine the Kinect's relative position.  See this example:

[Reply #3 on:]

You are probbly best exploring an ultra sonic solution with three or four external transducers pinging a receiver on your mobile device.

I thought of a system like that, but I didn't know what the device was called, do you know of a link or the name of a device like that?

If you move slowly enough your 3D scans should be highly correlated.  Perhaps some math would be enough to determine the Kinect's relative position.

So you mean something like camera tracking, with no external hardware? That would be great if I could figure out how.

[Reply #4 on:]

If this is an object scanner, does this mean that you can't rely on it being visible to a static camera? Putting a highly visible marker on it and tracking the position in 3D using video cameras would seem like your best bet to get accurate positions. I can't imagine any solution based on GPS or inertial movement sensing giving you anything like that accuracy. The only other way I can think of would be using a low powered RADAR-like range finder based on a reflector or transponder, but I can't think of any commercial applications that would provide a solution and it would be extraordinarily difficult to develop your own RADAR system (at all, let alone with this resolution).

[Reply #5 on:]

If you move slowly enough your 3D scans should be highly correlated.  Perhaps some math would be enough to determine the Kinect's relative position.

So you mean something like camera tracking, with no external hardware? That would be great if I could figure out how.

Good thing the comments on that YouTube video tell you the name of the technical paper: "RGB-D Mapping: Using Depth Cameras for Dense 3D Modeling of Indoor Environments".  Google found it here: http://www.cs.washington.edu/robotics/postscripts/3d-mapping-iser-10-final.pdf

[Reply #6 on:]

@PeterH Thats a good idea although I think it would take two or three normal cameras, although I could just use another kinect to get the position of the first one although I think im going to try johnwasser's idea first because it requires less componets.

@johnwasser Thanks for the link! Im going to try that, if that works it will be a simple solution to my problem in terms of cost and equipment.

Thanks everyone for the help