Mickey jitter

[GoogleCodeExporter]

What steps will reproduce the problem?
Use Mickey with TrackIR5 about 1 metre from (but not directly facing) a window. 
Tracking mode: absolute. ltr_gui model name "single point reflective" as I am 
using the circular sticker supplied for earlier TrackIR models.

What is the expected output? What do you see instead?
After a couple of minutes working fine there is severe jitter and then the 
pointer jumps to the left hand side of the screen and sticks there. This occurs 
even on very overcast days.

What version of the product are you 2. After a couple of minutes working fine 
there is severe jitter and then the pointer jumps to the left hand side of the 
screen and sticks there.
3. This occurs even on very overcast daysusing? On what operating system?
0.99.10 on Ubuntu 14.04 LTS 64 bit. 

Please provide any additional information below.
Processor: Intel® Core™2 Duo CPU P8600 @ 2.40GHz × 2 
8GB RAM
Not sure why the tracking device in ltr_gui shows as "wiimote" with no option 
to change it or whether that is anything to do with the problem.

Original issue reported on code.google.com by a...@geoff.kelsall.name on 4 Jun 2014 at 10:28

[GoogleCodeExporter]

Hello,
my bet is, that there is something competing with your reflective point.

To be sure you need to see what the tracker sees using ltr_gui; remember that 
TrackIR uses infrared light, so things like lit cigars, candles, lightbulbs, IR 
remote controls  and other things might make it go crazy...

You say the ltr_gui doesn't seem to be working right; may I ask you to attach 
here the log of the ltr_gui? (fourth pane, View logfile, if I'm not mistaken) 
It might shed some light on what is going on - if mickey sees the TrackIR, but 
ltr_gui does not, there is something strange happening.

Kind regards,

Michal

Original comment by f.jo...@email.cz on 5 Jun 2014 at 5:05

[GoogleCodeExporter]

Hello Michal,

Thanks for your quick reply. I attach the ltr_gui log but I cannot reproduce 
ltr_gui showing only "wiimote" so perhaps that was nothing to do with the 
problem. 

None of the items you mention above are in the room but you can see on the 
screenshot (there is a little sun today) that TrackIR has detected a couple of 
other points, which may be causing the jitter. When I look at the camera view 
it is always very obvious to me which is the correct point (the circular one 
slightly larger than the cross). I wonder if there are any settings of TrackIR 
or ltr_gui or mickey which would allow mickey to just pick that point out. If 
it is something that is easy for humans to pick out but difficult to program in 
software might an enhancement be to add a training mode in which I could follow 
the correct point manually for a while to train the software?

Regards

Geoff

Original comment by a...@geoff.kelsall.name on 5 Jun 2014 at 7:48

Attachments:

• ltr_gui_log.txt
• [Screenshot from 2014-06-05 08:27:22.png](https://storage.googleapis.com/google-code-attachments/linux-track/issue-67/comment-2/Screenshot from 2014-06-05 08:27:22.png)

[GoogleCodeExporter]

Hi,
there are several things that you can try...

First of all - quite obvious, but very often overlooked (been there, done that) 
is the possibility to change the camera's field of view; for example - you have 
a spot towards the bottom, while the correct marker is quite high in the 
picture. You can try to point the TrackIR a bit higher - your marker gets 
closer to the center (always better for a variety of reasons) and at least some 
of the noise might be out of play (provided there is not a lamp on the ceiling, 
or something alike).

You can also try to play with the threshold - try to see if it makes the wrong 
spots smaller...

Then there is the min blob setting - what it does is that it imposes a limit on 
minimum amount of pixels in a blob to be considered valid. Just be carefull, 
you could hit this limit with the marker too, should you get somewhat farther 
from the TrackIR camera...

And last, but not least - I see there is a small spot close to the marker; is 
that a reflection on glasses? It might be good to get rid of it, if possible. 
There you might try to either obscure the reflection somehow, or you can lower 
the IR illumination brightness (just remember, that you have to pause and 
restart the tracking to apply this change)...

Hope it helps...

Kind regards,

Michal

Original comment by f.jo...@email.cz on 6 Jun 2014 at 5:02

[GoogleCodeExporter]

Michal,

Doh I just bought expensive new glasses - guess I should have asked for a 
plastic frame.

Thanks for those suggestions - I will try them.

Can you give me update access to the wiki so I can add them and anything else 
useful I find to the virtual mouse doc?

Regards

Geoff

Original comment by a...@geoff.kelsall.name on 6 Jun 2014 at 7:58

[GoogleCodeExporter]

Hello Geoff,
if I may, I'd ask you to send the changes to me - the thing is that I'd like to 
have the changes done to the help in mickey/ltr_gui too and this way I'll be 
able to update both...

You can send them via mail, or I can create a specific "Docs update" bugreport 
thread to channel through these changes...

Thank you,

Michal

Original comment by f.jo...@email.cz on 6 Jun 2014 at 8:06

[GoogleCodeExporter]

I could not find your direct contact details so here is my suggestion for 
documentation update:

Troubleshooting
-----------------------
-----------------------

Mouse jitters or jumps when using infra-red tracking
---------------------------------------------------------------------------

In the ltr_gui tracking window start the camera view. You should see an image 
of the object being tracked, for example a circle if you are using a circular 
reflective sticker, with a target cross.

If you see multiple target crosses try to reposition the computer or other 
objects to avoid other infra-red emissions or reflections from for example lit 
cigars, candles, lightbulbs or IR remote controls. Windows may also cause 
interference: if so you may need to turn the tracking device away from the 
window and/or draw the curtains. Interference from spectacles may be resolved 
by changing the "valid blob size" as described below.

If you cannot identify the offending image in the camera view temporarily 
decrease the threshold to minimum in the device setup tab of the other ltr_gui 
window and check the camera view again.

Increase the threshold to maximum and if you cannot see the legitimate target 
image decrease it gradually until you can.

Change the "valid blob size" to "200 to 600" if you are using a standard 
circular target sticker or somewhat larger if you are using a larger target. 
The target cross will now disappear when you move the target very close and 
also when you move it very far away. Adjust the valid blob size to a range 
comfortable for you. 

Original comment by a...@geoff.kelsall.name on 6 Jun 2014 at 1:49