PDA

View Full Version : Flock Responses



Karen Reisiger
07-15-1999, 02:44 AM
Here are the responses I recieved from the posted Flock of Birds
accuracy problem. There
are some very good suggestions. Unfortunately, we are still
experiencing errors. We normalized the standard deviations with the
stylus length, and are recieving errors of .1 mm per 1 mm. This error
suggests to me that there may be a problem with our system. Again, any
other suggestions are welcome. Thanks for the great response.

Karen Reisiger
MSRC, University of Pittsburgh
(412) 648-1943


Without the stylus:

25.4 mm/inch * 0.07 inch RMS = 1.78 mm RMS

Maximum error possible:

3 max/RMS * 1.78 mm RMS = 5.33 max mm

(RMS is a statistical measure, approx. to Standard Deviation.
Maximum error should not exceed 3 time RMS)

The fact that you are getting 3-4 mm error with a stylus tell us that
your maximum error is much smaller, which is expected because
you are not moving over a great range.

Have you tried rotating the receiver around the receivers X axis
while calibrating. This may correct for any minor axis asymmetry. It
may make a difference, especially with numbers this small.

Given the large errors you are getting with your Flock (your biomch-l
posting of Jun 30), you should probably take
a look at

Nixon, MA, McCallum, BC, Fright, WR, Price, NB (1998) The effects of
metals and interfering fields on
electromagnetic trackers. Presence 7: 204-218.

if you haven't already. It compares Polhemus and Flock, and is a good
summary on the source of errors from these
devices. Especially note that the susceptibility of both of them to
noise increases as the 4th power of the
transmitter-receiver distance. Thus, this distance should be kept to a
minimum.

I just developed a stylus and tip estimation software myself. We get
errors of about 0.2 mm. However, my first attempt gave errors about 5
mm
as well. Where I went wrong was misunderstanding the M matrix given by
the
Flock. I was assuming it was the direction cosine matrix of the sensor
coordinate system on the transmitter system. It is the other way
around.
Therefore I should have been using the transpose of M in my
transformations. After I fixed that I got very small errors in
estimating
the location of the digitizing tip.

I am using the same system with a long stylus (500mm) to estimate a
point in space. I had a problems also in estimating the location of
the tip.

I am not sure if your are using the 'point' mode or the 'stream'
mode from the FOB. I am using the 'point' mode and found that it
takes the FOB a couple of tries before it works out where the sensor
is. By this I mean you have to send the ASCII command "B" about 3
times and before FOB accurately determines the location the sensor. I
don't know if this common to all FOB systems, but it certainly fixed
a lot of my problems. In stream mode this should not be a problem.

Also, as indicated in the manual, as the euler angle 'elevation'
approaches +-90 deg inaccuracies will occur. If your sensor
approaches these limits when estimating the centre of your sphere,
then errors will occur.

We had a Flock of Birds a couple of years ago. Aside from having a
really bad room for any kind of electromagnetic tracking device, we
noticed that we could fairly easily induce artifact in signals from
the automatic gain control. If the distance from the transmitter to
receiver is changing appreciably during your test, this may be the
problem. The solution is to orient the subject to the transmitter in
such a way that the distance from receiver to marker is as "constant"
as possible.

this is what we experienced too. It appeared to be necessary to
calibrate
the Bird system for the specific environment in which you are measuring.

Last year a paper on this problem was published in Clinical Biomechanics

(13; 280-292). Should you want more information on the algorhythms, send

a
mail to Jochem Nagels (jnagels@ortho.azl.nl).

We also have the Ascension Flock-of-Birds EM tracking device and have
performed electromagnetic interference (EMI) testing on the equipment.
We
have determined that the equipment is susceptible to magnetic fields
greater than 2.6 mG at 60 Hz. You may want to check the environment in
which you are using the equipment to be certain that you are not picking

up
EMI.


We had been using the device for a long time. Magnetic trackers extract
postion
and orientation from 9 parameters, so by the construction, the
positional error
depends on orientation (and vice-versa). We noticed significant error
while we
rotate the sensor. 5-10mm difference isn't very surprising. (There may
be other
problems, too...) If you keep the orientation of the sensor and move
along a
ruler 5cm and you still get 5mm-10mm error relative to the distance you
moved, I
can tell something must be very wrong.
I have an impression that Polhemus is much better.
Optical trackers we use now (FlashPoint) is way way better.


Be sure that you rotate around the receiver's X axis while taking
data for the sphere.

As I understand it, you are calculating the mean X,Y, and Z offset
from these values and then using this to generate the tip location
when the receiver is moving unconstrained.

There will be some variability in the offset distances which will
reflect
the non-orthogonality of the receiver coils.

After taking the mean distances - run the test again to see the offset.
If it continues to be high we should plan on having the tracking
recalibrated here at Ascension. I wonder if you have the latest
firmware and engineering upgrade. (what is the serial number (SN)
listed on the back of the Flock of Birds?)


Make sure you take off your watch and any rings or bracelets that you
might
be wearing.

As well, make sure that your stylus tip is inside a hole that is as
small
as possible and not very deep. Lift-off errors are common with
unnecessarily large or deep holes. When taking data, do not make
exceed
an arc of about 90 degrees or so from vertical. You might think that
the
larger the arc, the more accurate the result, but this is not the case,
because it becomes theoretically impossible to maintain good placement
of
the tip in the hole with greater and greater stylus angles.

Make sure the point and body of the stylus is made of a very hard
material. Some poly materials are too flexible for this application.

Finally, a little know fact about the Flock of Birds is that it relies
on
the signal form all the birds to determine its field strength/accuracy
during data collection. Make sure that the other birds are no farther
or
closer to the source than the bird on the stylus during data collection.

It uses the furthest bird to calibrate its field strength and thus its
accuracy.

Increase your distance from the computer and make sure that the table
doesn't have any metal under it.

I hope this helps, and remember, don't forget the wrist watch!

the things that i would consider are:

(1) the mode in which you are collecting the angular information from
the
birds. The cardan angles used in the angles mode will give you
undefined
readings at 90 degrees of elevation (an inherent property of cardan
angle
systems). If you need to collect angular information through more than
90
degrees of elevation, consider using quaterions or matrix mode.

(2) the transformation that you are using to get from the sensor to the
stylus tip. Just one sign reversal or sine for cosine substitution will

cause the sort of errors you are getting.

(3) It does not sound like a problem with bouncing of the magnetic
field,
altough 2-d (i.e. sheets) of metal even 5 feet away can cause problems
if
you have the receiver far from the transmitter (as you move the reciever

away from the transmitter, the strength of the magnetic field increases,

increasing the cahnce of bouncing)

(4) if none of these are the problem, call ascention and talk to steve
work,
he is the head of tech support and is a great guy to work with.

---------------------------------------------------------------
To unsubscribe send SIGNOFF BIOMCH-L to LISTSERV@nic.surfnet.nl
For information and archives: http://isb.ri.ccf.org/biomch-l
---------------------------------------------------------------