Using Digital images to count fish in water past a particular point

[ozfish]

Can anybody tell me how I could use a digital video camera to count fish going through a channel (say 300mm wide) Note the fish are of various sizes and may swim in groups.

I look forward to replies

 

[dydt]

fish are all same size?

its possible to use a phoelectric sensor tied to a counter...can you limit the size of the channel so that they will travel one at a time?

dydt

 

[busbar]

Would the concept of electronic pill counters be applicable?



Epilogue... Search 'Drug-O-Matic' 'Health Care Logistics' also

http://www.scguild.com/Resume/2174R.html

string 'Pill counter using 8051 processor'

 

[Cappels]

Fule for thought: If you know enough about the fish you will be counting, you might want to estimate the mass of the fish by imaging from two orthogonal positions (top and side camers), then apply statistical methods base on the distribution of volume of the (X percent are y cc, m percent are n cc, etc. and they have a packing factor of Z) fish to determine how many went through.

This might provide more accurate results than trying to actually count them.

 

[BJC]

Have you contacted the Corps of Engineers at Bonneville Dam? They have a camera in place that counts fish. How well I don't know because they sometimes have a person there also.

 

[busbar]

On BJC's sugestion...a couple of long shots:

US Army Corps Of Engineers, The Dalles, OR 97058 (541) 296-1181

Another possibility is Bonneville Power

http://www.bpa.gov

 

[busbar]

http://www.usace.army.mil/faq.html

 

[jbartos]

Suggestion. It appears that camera alone would not be accurate during the night. You would also need a light fixture.
It may be a good idea to explore other principles, e.g. ultrasonic systems. Also, notice that you will probably need a vertical and horizontal arrays of sensors since the fishes can pass side by side as well as one on the top of the other. The the array output signals will have to be processed by logic circuits to obtain accurate fish count.

 

[xray]

Hey Ozfish:

I am gonna through in my two bits and make it swim. I recently finished a excercise that utilized a 1N4148 DC amplified and sent to an on chip ADC. I think with the same theory getting enough of these in a line and mounting them to a flat screen video output say a TV. Fish pass the line of diodes creating enough continuos fish signal implementing Cappels theory, orthogonal views at the same position, two signals one fish mass. Orthogally opposed signal could be analyzed by statisistics to give a three dimensional accurate count.
With that said, It seems like overkill. I think a couple of ultrasonic sensors set up in the right array sending signal to an ADC on a chip and some digital counting circuitry would yield a more accurate count for less.

Happy counting to you.

Cy Drollinger

 

[jbartos]

Suggestion: Visit

http://www.uesystems.com/techview.htm

http://www.hpsensors.com/success.asp

http://www.massa.com/

etc. for more info

 

[chakorules]

Whoa...that would be a neat VISION APPLICATION! But sounds like it would be very tough. Lighting would be very important. I would recommend a high end vision system like Acuity

http://www.rvsi.com/acuitycimatrix/index.htm

or DVT

http://www.dvtsensors.com/products/

If you can take a digital picture of the fish, you can download DVT's Framework from their website and see if you can detect various fish with different JPG images you take and load them into the FrameWork software with little or no money you could detrimine if this type of vision system will work for your vision "Fish" project....sounds kewl! Chris Elston
Automation & Controls Engineer

http://www.mrplc.com

Download Sample PLC Ladder Logic Code
at MrPLC.com

 

[jbartos]

Suggestion: It is also necessary to consider the water quality; especially pollution, e.g. clay, leaves, etc. Leaves, diftwoods, etc. could potentially be counted as fishes.