About aquaculture fish counting

[yuanyuan]

Hi to all, Have you thought about this as well as for me? in fish farming the fish are fed at a percentage of their weight. i measured the weight of the fish that i put into the tank, and i can remove the fish once in a while to reweigh them but that is very stressful to them. i would like to insert a laser into my fish tank, that can count the amount of time that the beam is interrupted by a fish- kind of like the lasers that are at the entrances to stores. i think there should be a strong correlation between the amount of time that the laser beam is covered and the weight of the fish in the tank.

First, we know water is pretty much non-absorbing in the visible range (which is why water appears clear to the eye). So, any visible laser will work decently, assuming your water is pretty clean. If you'd like to be precise about it, here's the absorbance spectra of water:

The big dip in the center corresponds pretty closely to the range of visible light, and we note that the minimum absorbance is around 400-500nm. So, pick up a laser diode such as this (out of stock atm, but you can find one on ebay or whatever easily), which is 532nm (close enough). Then, grab a photosensitive receptor (like this photocell, which happens to have a max sensitivity at around 500nm - but a phototransistor or any similar sensor would work also). Put the laser at one end of your tank, shooting through a clear window, and have the sensor at the other end acting as a detector. Hook up a simple MOSFET driver circuit for the laser http://www.kynix.com/Detail/457952/LASER.html, and read the voltage off the photocell on an analog pin - figure out where a good analog cutoff for on/off is, and you're pretty much set on the hardware side.

I expect the harder part of this project will be getting a good correspondence between the laser signal and fish weight - but if you collect enough data to get a decently smooth laser % time vs fish weight, you should be all set. Obviously you'll want to get this data by recording the laser data for a growing cycle, and manually annotating the graph with the fish weights - so you can try to get some sort of best-fit curve to use for future predictions.

[Bas Pels]

I wonder whether this is serious. However, I will give you the bvenefit of the doubt

What a laser here does, what any laser can do, is determine wheter there is something between sender en collector. This could be a fish, or it could be, for instance, a floating leaf. The laser will not be able to determine what went along.

Whether the obstruction is a large, fat fish, or a small one, which happens to take a break from swimming in the laser light will not be visible. This would require assumptions. I can imagine you will exclude the coincidences I gave you, but than what the laser might do is provide you with bodylength * swimmingspeed information.

In most fish species swimmingspeed is correlated with bodylength, so basically you get bodylength square * a factor. Assuming the fish Always swim in the same direction. Which they generally don't. But perhaps statistics can rule this problem out.

But bodylength does not equal weight. Assuming a fish does not change shape (and in many cases this is an incoprrect assumption) the weight will equal lenth to the third power times another factor.

That is, if a fish does not change its shape, it will have to be twice as thick, and twich as high in case its length will become double.

I think it would be better to firstly determine bodyweight as a function of bodylength.

So that, if you know there is 1000 fish of 3 cm length, you know how much they weigh.

Another thing is, I once took a course of fish farming, and the thing is, a small fish does need more food (per kilo) than a big one does. If you want to draw a line through the masspoints, you will have to take the weight to the power of 0.8

But his can be standarized too: 1000 fish of 3 cm means still a certain amount of food. And if they were 800 of 5 cm, the amount would be another one

[Lycosid]

I've used beam-break sensors for behavioral monitoring and even estimating counts in populations. However, the number of stacked assumptions here worries me.
First, your number will be good for one species of fish in exactly that tank. In some other tank their activity might be different and the number wouldn't be good. If you are running three tanks you won't be able to combine your data to give you a larger sample size (and you'll need a huge sample size).
Second, behavior of the fish may change across ontogeny. If fish at size X swim many body-lengths per minute but fish at length 4X swim at half a body length per minute and sometimes flat out stop and hover in your beam the difference in beam coverage is going to be totally non-linear.
Third, because this is behaviorally variable there could be day-to-day or season-to-season changes. The fish all become active because extra sunlight is hitting the tank and your number that day goes completely off the rails.

Now, if you were running this as a data-collection method in one location for 15 years I'd guess that you could average through those errors and get some sort of useful trend. However, for deciding how much food goes in the tank this week I don't see this providing numbers with small enough error margins to be safe.

You MIGHT be able to do better with a network of pressure sensors since the pressure output of a swimming fish is directly tied to muscle mass, but that would only have any chance of working in areas where nothing else disturbs the water. And I'm not sure that "better" is "good enough".

[yuanyuan]

I wonder whether this is serious. However, I will give you the bvenefit of the doubt

What a laser here does, what any laser can do, is determine wheter there is something between sender en collector. This could be a fish, or it could be, for instance, a floating leaf. The laser will not be able to determine what went along.

Whether the obstruction is a large, fat fish, or a small one, which happens to take a break from swimming in the laser light will not be visible. This would require assumptions. I can imagine you will exclude the coincidences I gave you, but than what the laser might do is provide you with bodylength * swimmingspeed information.

In most fish species swimmingspeed is correlated with bodylength, so basically you get bodylength square * a factor. Assuming the fish Always swim in the same direction. Which they generally don't. But perhaps statistics can rule this problem out.

But bodylength does not equal weight. Assuming a fish does not change shape (and in many cases this is an incoprrect assumption) the weight will equal lenth to the third power times another factor.

That is, if a fish does not change its shape, it will have to be twice as thick, and twich as high in case its length will become double.

I think it would be better to firstly determine bodyweight as a function of bodylength.

So that, if you know there is 1000 fish of 3 cm length, you know how much they weigh.

Another thing is, I once took a course of fish farming, and the thing is, a small fish does need more food (per kilo) than a big one does. If you want to draw a line through the masspoints, you will have to take the weight to the power of 0.8

But his can be standarized too: 1000 fish of 3 cm means still a certain amount of food. And if they were 800 of 5 cm, the amount would be another one

Thanks for your reply and suggestion.

[TwoTankAmin]

While it is stressful handling fish, I wonder about a different approach. If you have 50 fish in a tank and you have a net basically as wide and tall as the tank, you can remove all the fish fairly easily. If you then have a large container on a scale with a known weight of water in it into which you dump the net full of fish, you can get an average weight pretty fast and then pour all the fish back into the tank. I am not sure how badly stressful this might be, especially if you are only doing it once a week.