search

twVolc / PyCamPermanent

Permanent PiCam (SO2) installation project software
GNU General Public License v3.0
1 stars 2 forks source link

Running the "histogram" fit with optical flow slows down if NaNs are detected along the plume cross section #61

Open tpering opened 1 year ago

tpering commented 1 year ago

I think the above is correct, next time I come across the error I will copy the error which occurs. The slow down is quite significant. Not a priority for real-time processing which I think will work better with just cross-correlation for the time being.

twVolc commented 1 year ago

I'm also experiencing pretty significant slow-down trying to process an entire day without breaking it up into chunks. I'm not 100% sure it's the histogram NaNs that are causing the problem as earlier in the sequence it processes them pretty quickly even when throwing NaN warnings, and vice versa it still seems quite slow when processing images that don't throw the NaN warnings later in the sequence. Possibly things are clogging up - memory issues - it would definitely be nice if it can be sped up as it's painfully slow at about 6 hours into a sequence!

twVolc commented 1 year ago

Just to note that my processing runs much faster right to the end of a day's time series when using my new option of a preloaded calibration. So that makes me think that it's something to do with the DOAS calibration and iteratively adding/removing points which starts to slow things down. Or this is at least part of the problem.