Detecting particles deformation /merging in a time lapse

I am trying to establish a systematic way to detect in which image of a time lapse particles deform /merge. The particles are plastic pellets, and I am aiming at determining the melting point, related to the time. The particles move due to buoyancy, and the melting can be visualized by deformation of the pellets followed by merge. Here the first and last imaged of the secuence, DSC_8425.tif (536.9 KB)

DSC_8780.tif (536.9 KB)

A video of the full secuence can be seen here Dropbox - Autoclave_meltingT_testtest.mpg - Simplify your life


While I can visualize deformation, since the pellets are not circles but ovals, I don’t find a geometrical parameter which I could use as a threshold. I welcome ideas. Otherwise I guess I need to go to the particle merge event. I have explored Track Mate with FiJi, but I have troubles in the initial particle detection stage as the particles are not circles. Are there any work-arounds?
I am now exploring Cell Profiler and the Object Tracking sample pipeline. While I’m still working on the correct particle detection, I would appreciate feedback on if I’m going in the right direction at all. Should I process the secuence backwards, to detect when the final blob splits rather than when it merges?
Thanks in advance!

Hi @Luci ,

I checked your movie. I think you have started fine and proceed in the same direction.
Here is the workflow that I can think of,

  1. illumination correction to remove unwanted light from your image.
  2. Identifying the particle (your oval object). For thresholding you need a thresholding value & rough size of your object. For example below is the screenshot from your movie when I have outlined your 5 objects/particle. This you could try this for all the frames in your movie. The later frames will be a big object as shown below.
  3. Once you identify the object measure the AreaSize using MeasureObjectSizeShape which includes Area & circularity as the parameter. For example, number of objects & area would tell you if they are multiple objects or single object.

You can give a try & if possible share your pipeline with sample image we can help with thresholding if needed.

Hope this helps,



Hi Lakshmi,

Many thanks for your feedback!

I’m on the thresholding step, indeed it identifies the light through the vial as particles, but I will work on it and share my pipeline as soon as it’s medium decent.


Hi @lakshmi, hi all!

so during these weeks I have been working both on the pipeline and on the design of experients, and gone to placing only one particle, there is then no need to track but just threshold and measure the particle itself, which is straightforwards. I’ve got my pipeline in place which is to a big extent working as should, but in some frames the threshold doesn’t do such a nice job, distorting my results. It detects artifacts in the background, although I have checked that the intensity does not change in those regions. I have followed the help suggestion to measure the backgrund intensity and add it as lower bund threshold, but it doesn’t remove the issue.
I have modified the smoothing scale and threshold correction factor, but don’t manage to improve threshold. Would be greatful for some tips!
Threshold example4 threshold6 Threshold8 [pipeline_MeltingT_v2.cppipe|attachment]
(upload://jdD9gvrj39H0wFoNmSfoGNZUSZu.cppipe) (12.7 KB)

PS Not sure if I should start a new thread since it is no longer about merging particles?

Hi! I think I’m good with the thresholding now, changing from thresholding strategy adaptative to global did the trick. Thanks!

Hi @Luci,