Is this something you would be interested in?

http://riverwatcher.is/

Automatic fish-counters. Originally developed for Atlantic salmon in Icelandic rivers but I think the have expanded it to other species.

The automatic counters may work, but I would think current, underwater vegetation, changing light conditions would make motion sensing cameras difficult to use.  You would probably want to consider some of the video analysis software that is available to eliminate 'dead time' if using video monitoring.

Do you have a model of camera in mind or are you wondering if such a camera exists? 

The system we recently develop may be of interest for you: http://onlinelibrary.wiley.com/doi/10.1002/rra.2984/abstract

Best regards

Dom

depends on the nature of the artificial passageway, but in general, I would expect you to get a lot of false positive video clips if there's any cavitation (bubbles) or vegetation (either leaves/sticks going through or any filamentous algae) in the frame.  Another option would be to write a script to analyze the video data and filter out the dead space.

Dominique, can I possibly get a copy of your paper please?

Annette, thanks for your interest.  I sent you the pdf file privately.

Regards

Did you find a suitable setup? I am looking for something similar. Kresimir Williams has a paper at: https://www.researchgate.net/publication/272568814

I think it's same to Crowd Counting tech with machine learning https://www.analyticsvidhya.com/blog/2019/02/building-crowd-counting-model-python/