search

soft-matter / trackpy

Python particle tracking toolkit
http://soft-matter.github.io/trackpy
Other
443 stars 131 forks source link

Question about sub pixel results #311

Open Resonanz opened 8 years ago

Resonanz commented 8 years ago

Hi guys, I have used the sub pixel code as per the walkthrough and have indeed found that there is a dip at the center of the histogram if the diameter setting is too small. That all good, but what I have found for all my experiments so far is a peak at the center. Have you seen this? Any idea of the reason for it? FWIW I have been using 500 nm particles and a diameter setting of 13.

screen shot 2015-11-07 at 10 28 07 am
danielballan commented 8 years ago

Interesting! Two questions:

  1. What version of trackpy are you using? Check tp.__version__.
  2. Are you able to share a sample image if anyone has time to play with this? (I might!)
Resonanz commented 8 years ago

In answer to the first question..... 0.2.4.

For the second, is there a way to upload a .tif file and share it? I can look into this later today, or you could tell me really quickly :-)

danielballan commented 8 years ago

GitHub lets you do image attachments -- maybe arbitrary attachments too? Not 100% sure that they accept TIFF.

On Mon, Nov 9, 2015 at 7:05 PM Alpha Particle notifications@github.com wrote:

In answer to the first question..... 0.2.4.

For the second, is there a way to upload a .tif file and share it? I can look into this later today, or you could tell me really quickly :-)

— Reply to this email directly or view it on GitHub https://github.com/soft-matter/trackpy/issues/311#issuecomment-155239021 .

tacaswell commented 8 years ago

I have a vague memory of seeing things like this before (mind you this was 4+ years ago now (!?)) where if you use a too big mask and have dense particles you get a bias towards the center due to the edges of the CoM mask picking up weight from the neighboring particles. Because they are farther out they have greater effect per count on the CoM and if you have lots of lines of particles you will get this bias as the wings will a) dominate and b) balance each other.

Also check that you are not finding false particles in the empty spaces. If you get a small local maximum in a region that is basically constant, the CoM will be centered on the pixel with the local max.

caspervdw commented 7 years ago

Probably too late, but this might have to do with smoothing_size. See #401.