[time-nuts] position determination over short distance

EB4APL eb4apl at cembreros.jazztel.es
Mon Nov 24 14:20:02 UTC 2008


Of course you can use any of the standard photogrammetric tecniques and 
programs.  Even you can do it automatically using image correlation to 
find the paralax and hence the distance.

But I have a simpler and cheaper (and also accurate) solution:

Put a transmiting coil in the object to be measured, about 1-2 cm 
diameter, 0.5 cm high (I don´t remember the frequency now, it is quite 
low but it doesn't matter).
Lay a receiving antenna grid of vertical and horizontal wires, spaced 
about 1 cm, in a surface below the object.
Scan the antennas noting the amplitudes an phases on the 4 or 6 
horizontal and vertical wires surrounding the Tx coil in each moment .
Calculate the position in real time (you obtain a precision of about 0.1 
mm if the antenna surface is calibrated)
Sound complicated? buy a digitizing tablet (used ones are really 
cheap).  This is how these things works.
Regards,
Ignacio Cembreros

Lux, James P wrote:
> There's a fair amount of F/OSS software from JPL available to do this sort of calibration. It's used to calibrate cameras used on Mars rovers, among other things. The target pattern for calibration is a bunch of big circular dots on a background.
>
>
> On 11/23/08 7:56 PM, "Bruce Griffiths" <bruce.griffiths at xtra.co.nz> wrote:
>
> Tom Van Baak wrote:
>   
>>> I need to determine the position of a instrument with a 1mm accuracy or less.
>>> The instrument is not connected to a mechanical device but is separate &
>>> independent.
>>> The surface which the instrument is positioned on is close to the size of a
>>> 11"x11" square.
>>>
>>>       
>> 1 mm or better accuracy on a 300x300 mm surface can be
>> obtained with a cheap webcam mounted above the surface
>> and a little bit a creative software.
>>
>> /tvb
>>
>>     
> Dont forget to calibrate the camera distortion and ensure that this
> doesnt over over time.
>
> i.e. lock the focus and ensure the camera position, tilt etc with
> respect to the reference surface do not vary.
>
> Bruce
>   




More information about the time-nuts mailing list