search

soft-matter / trackpy

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

trackpy.locate returns clarification #666

Open b-grimaud opened 3 years ago

b-grimaud commented 3 years ago

I am using trackpy.locate (or rather, .batch) and then .link to form trajectories, which are then processed even further. This includes calculating the volume of the 2D gaussian for each feature found, which made me wonder what exactly is the meaning of the mass return of .locate. It is described as mass means total integrated brightness of the blob, but that data is sometimes orders of magnitude smaller than the integral I calculate separately. Furthermore, the signal return isn't actually described anywhere. It might seem self-explanatory, but trackpy.locate's documentation skips over it. I know this issue has previously been mentionned here #625 and here #483, but I never found a clear explanation of what it returns.

nkeim commented 3 years ago

Thanks for bringing this up. signal is the signal-to-noise ratio S/N that is discussed in the uncertainty estimation tutorial. Please keep this issue open so we can improve the documentation!

weidtn commented 3 years ago

Also can you clarify what is ep?

nkeim commented 3 years ago

ep is also described in the uncertainty estimation tutorial. It estimates the uncertainty in the particle's position, by the method of Savin and Doyle.

@BptGrm I didn't see earlier that you also asked about mass. I'm not quite sure why you would get such a large discrepancy, but it may be due to image preprocessing, or conversion from an integer image (normalized to 255) to float (normalized to 1.0). Ordinarily the mass is the sum of pixel intensities within the given feature diameter, and it is computed from the raw_image supplied to locate().