Window size & target resolution

amirdastgheib commented 7 months ago

How did you install pyopenrivercam (mamba / conda or pip)

pip

What is the problem?

Hi I have a problem with selecting the optimal values for window_size and target resolution. My river width is around 50 m and the discharge is 130 (m3/s) so, the real flow velocity is around 2(m/s) in reality. I have used OpenRiverCam and I selected window_size = 70 with different target resolution = 0.05, 0.06, 0.07, 0.08 and for each target resolution the mean discharge (quantile 2) shows different value. and the same for changing window_size with fix target resolution. However, changing mean v_eff is not that much and is from 2.2 to 2.4. I am not sure which window_size and target resolution would be the optimum values for this river. Also, why when I changed the window_size and target resolution, discharge shows more fluctuations in comparison with v_eff??

How can we reproduce this? (What steps trigger the problem? What parameters are you using for processing? Include screenshots. If you are having issues processing a video, if possible include a copy of your dataset uploaded on Dropbox, Google Drive, or other)

[Type answer here]

amirdastgheib commented 7 months ago

Window Size_ Resolution

amirdastgheib commented 7 months ago

Also I did it with another water level to see the difference: Window Size_ Resolution 2

amirdastgheib commented 7 months ago

As I said, the real discharge is 130 (m3/s) and according to our cross section the real velocity would be around 2 (m/s).

hcwinsemius commented 7 months ago

I went through your table. Looks like you did comprehensive investigating. A few remarks:

a window size of 70 with a resolution of 0.1 m is a box of 7x7 meters. That is likely much too high to properly track features. You may need a smaller window size. I tend to use 20 or so.
if you resample to 0.1 m or below you probably loose a lot of features in the resampling. Try to stay at or below 0.05m
the variability in velocities can be caused by many things. Important ones are variability in frame rates that are not captured by the device. But also natural variability in turbulence or (if hand-held) instability in the image objective.
probably you are doing this: also make plots of the situation using our highly convenient plotting methods on top of the Velocimetry and Transect classes. This may reveal a lot of things
we have a new (v0.5.3) project method that resolves issues with objectives that are partly outside the field of view which can cause errors in the reprojection. Try this out with frames.project(method="numpy") or (if using CLI)
```
frames:
....<whatever preprocessing steps you have>
project:
    method: numpy
```
We are always very curious to see a result from your analysis.

amirdastgheib commented 7 months ago

Thank you for your informative response. According to your last point: "we have a new (v0.5.3) project method that resolves issues with objectives that are partly outside the field of view which can cause errors in the reprojection. Try this out with frames.project(method="numpy") " I am a little bit confused. Indeed, I am not sure if I can do (frames.project(method="numpy")) in my current version (v0.5.2) or I need to first install new version (v0.5.3) and then do (frames.project(method="numpy"))?? And if so, how I can install v0.5.3?

amirdastgheib commented 7 months ago

And if I can do (frames.project(method="numpy")) with my current version (v0.5.2), I am not sure where should I put (frames.project(method="numpy")) to my code. Because in several lines we have (frames.project()) such as:

da_norm_proj = da_norm.frames.project() or

da_rgb_proj = da_rgb.frames.project() in (Analyze surface velocities of a video with velocimetry) or da_rgb_proj = da_rgb.frames.project() in ( Immersive plotting and analyzing results)??? I did not get where exactly I should put this : frames.project(method="numpy")!

hcwinsemius commented 7 months ago

In any line where you perform projection you can use method=numpy

Updating to the latest code with conda:

conda update pyopenrivercam

With pip:

pip install --upgrade pyopenrivercam

localdevices / pyorc