search

ivoflipse / Pawlabeling

Tool for processing and analyzing pressure measurements
Other
18 stars 1 forks source link

Incorrect invalid contacts #3

Closed kreupeldier closed 11 years ago

kreupeldier commented 11 years ago

I have several measurements where paws in the middle of the plate get labeled as invalid. I suspect this is due to the force threshold at the start and end of a contact. Please a have look what would be a more reliable value for this!

incorrect invalid paws

ivoflipse commented 11 years ago

I'm seeing multiple examples where it seems the plate starts to measure too late, which causes the first visible frame of a paw to be exaggerated.

Frame 0 of measurement + frame 0 of contact image

Frame 92 of measurement + frame 0 of contact image

Notice how light the paw touches the plate in contrast with the first step?

The values clearly show this too. Where regular steps have a total force of 2-5N on their first frame and a max of 100-250N, the first two steps have a total force of 80 and 110N for the first frame with a max of 240 and 110N respectively.

While I think it depends on how the dog puts his paws on the plate, clearly the threshold can't be too high either. Though it shouldn't filter out paws in the middle of the measurement. Perhaps I should calculate a percentage for all paws and then use the average + 1 or 2 SD to determine a cut-off with a certain maximum bound.

ivoflipse commented 11 years ago

I've set the threshold to 0.25 * maximum sum of the force over time. If there are any examples where this doesn't seem to work well enough, I'll look into it again.